model.json

{
  "modelTopology": {
    "node": [
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "ConstantFolding/Postprocessor/Decode/div_recip",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1917"
                  },
                  {
                    "size": "4"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "Postprocessor/ExpandDims",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "ConstantFolding/Postprocessor/Decode/div_4_recip",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 3
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "2"
                  }
                ]
              }
            }
          }
        },
        "name": "Postprocessor/Reshape_1/shape",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "ConstantFolding/Postprocessor/Decode/div_2_recip",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "12"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "12"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "1280"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "1280"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "512"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "512"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                1917
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "MultipleGridAnchorGenerator/assert_equal/x",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                4
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_0/stack/3",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                2
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/ExpandDims_1/dim",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "273"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/ClassPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "273"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_0/ClassPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                1083
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_0/stack/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "1280"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "1280"
                  },
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                600
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_1/stack/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "512"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "512"
                  },
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                150
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_2/stack/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                54
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_3/stack/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                24
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_4/stack/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/gamma",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/beta",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/moving_mean",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/moving_variance",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "546"
                  }
                ]
              }
            }
          }
        },
        "name": "BoxPredictor_5/ClassPredictor/weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "546"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor/biases",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "160"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "160"
                  },
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "960"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "960"
                  },
                  {
                    "size": "160"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "160"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "160"
                  },
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "960"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "960"
                  },
                  {
                    "size": "160"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "96"
                  },
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "96"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "96"
                  },
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "64"
                  },
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "64"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "64"
                  },
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "64"
                  },
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "32"
                  },
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "192"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "192"
                  },
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "32"
                  },
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "192"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "192"
                  },
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "24"
                  },
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "144"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "144"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": []
              }
            }
          }
        },
        "name": "Preprocessor/mul/x",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                0
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/TensorArrayStack/range/start",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "Preprocessor/sub/y",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "32"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "32"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "32"
                  },
                  {
                    "size": "16"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "16"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "16"
                  },
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "96"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "96"
                  },
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "24"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "24"
                  },
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "144"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "144"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "144"
                  },
                  {
                    "size": "32"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "32"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "32"
                  },
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "192"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "192"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "192"
                  },
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "64"
                  },
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "384"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "384"
                  },
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "96"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "96"
                  },
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "576"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "576"
                  },
                  {
                    "size": "160"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "160"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "160"
                  },
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "960"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "960"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "960"
                  },
                  {
                    "size": "320"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/project/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "320"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "320"
                  },
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv_1/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1280"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "1280"
                  },
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "512"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "512"
                  },
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "128"
                  },
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "256"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "256"
                  },
                  {
                    "size": "64"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  },
                  {
                    "size": "3"
                  },
                  {
                    "size": "64"
                  },
                  {
                    "size": "1"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise_weights",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 1
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/mul",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "64"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  },
                  {
                    "size": "1"
                  },
                  {
                    "size": "64"
                  },
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Conv2D/merged_input",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 1,
              "tensorShape": {
                "dim": [
                  {
                    "size": "128"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/sub",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                0
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/strided_slice/stack",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                1
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "1"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/strided_slice/stack_1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                6
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_5/stack/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                91
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "BoxPredictor_0/stack_1/2",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 3
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                1
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          }
        },
        "name": "Postprocessor/Tile/multiples/1",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/Slice/begin",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "3"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/Slice/size",
        "op": "Const"
      },
      {
        "input": [],
        "attr": {
          "dtype": {
            "type": 4
          },
          "shape": {
            "shape": {
              "dim": [
                {
                  "size": "-1"
                },
                {
                  "size": "-1"
                },
                {
                  "size": "-1"
                },
                {
                  "size": "3"
                }
              ]
            }
          }
        },
        "name": "image_tensor",
        "op": "Placeholder"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayStack/range/start"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "is_constant": {
            "b": false
          },
          "parallel_iterations": {
            "i": "32"
          },
          "frame_name": {
            "s": [
              80,
              114,
              101,
              112,
              114,
              111,
              99,
              101,
              115,
              115,
              111,
              114,
              47,
              109,
              97,
              112,
              47,
              119,
              104,
              105,
              108,
              101,
              47,
              119,
              104,
              105,
              108,
              101,
              95,
              99,
              111,
              110,
              116,
              101,
              120,
              116
            ]
          }
        },
        "name": "Preprocessor/map/while/Enter",
        "op": "Enter"
      },
      {
        "input": [
          "image_tensor"
        ],
        "attr": {
          "SrcT": {
            "type": 4
          },
          "DstT": {
            "type": 1
          }
        },
        "name": "ToFloat",
        "op": "Cast"
      },
      {
        "input": [
          "Preprocessor/map/while/Enter",
          "Preprocessor/map/while/NextIteration"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "N": {
            "i": "2"
          }
        },
        "name": "Preprocessor/map/while/Merge",
        "op": "Merge"
      },
      {
        "input": [
          "ToFloat"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/TensorArrayUnstack/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayUnstack/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "Index": {
            "type": 3
          },
          "shrink_axis_mask": {
            "i": "1"
          },
          "begin_mask": {
            "i": "0"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          }
        },
        "name": "Preprocessor/map/TensorArrayUnstack/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayUnstack/strided_slice"
        ],
        "attr": {
          "dtype": {
            "type": 1
          },
          "element_shape": {
            "shape": {
              "dim": [],
              "unknownRank": true
            }
          },
          "dynamic_size": {
            "b": false
          },
          "clear_after_read": {
            "b": true
          },
          "identical_element_shapes": {
            "b": false
          },
          "tensor_array_name": {
            "s": []
          }
        },
        "name": "Preprocessor/map/TensorArray_1",
        "op": "TensorArrayV3"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayStack/range/start",
          "Preprocessor/map/TensorArrayUnstack/strided_slice",
          "Postprocessor/Tile/multiples/1"
        ],
        "attr": {
          "Tidx": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/TensorArrayUnstack/range",
        "op": "Range"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayUnstack/strided_slice"
        ],
        "attr": {
          "tensor_array_name": {
            "s": []
          },
          "dtype": {
            "type": 1
          },
          "element_shape": {
            "shape": {
              "dim": [],
              "unknownRank": true
            }
          },
          "dynamic_size": {
            "b": false
          },
          "clear_after_read": {
            "b": true
          },
          "identical_element_shapes": {
            "b": false
          }
        },
        "name": "Preprocessor/map/TensorArray",
        "op": "TensorArrayV3"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayUnstack/strided_slice"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "is_constant": {
            "b": true
          },
          "parallel_iterations": {
            "i": "32"
          },
          "frame_name": {
            "s": [
              80,
              114,
              101,
              112,
              114,
              111,
              99,
              101,
              115,
              115,
              111,
              114,
              47,
              109,
              97,
              112,
              47,
              119,
              104,
              105,
              108,
              101,
              47,
              119,
              104,
              105,
              108,
              101,
              95,
              99,
              111,
              110,
              116,
              101,
              120,
              116
            ]
          }
        },
        "name": "Preprocessor/map/while/Less/Enter",
        "op": "Enter"
      },
      {
        "input": [
          "Preprocessor/map/TensorArray_1:1"
        ],
        "attr": {
          "is_constant": {
            "b": false
          },
          "parallel_iterations": {
            "i": "32"
          },
          "frame_name": {
            "s": [
              80,
              114,
              101,
              112,
              114,
              111,
              99,
              101,
              115,
              115,
              111,
              114,
              47,
              109,
              97,
              112,
              47,
              119,
              104,
              105,
              108,
              101,
              47,
              119,
              104,
              105,
              108,
              101,
              95,
              99,
              111,
              110,
              116,
              101,
              120,
              116
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/Enter_1",
        "op": "Enter"
      },
      {
        "input": [
          "Preprocessor/map/TensorArray_1"
        ],
        "attr": {
          "frame_name": {
            "s": [
              80,
              114,
              101,
              112,
              114,
              111,
              99,
              101,
              115,
              115,
              111,
              114,
              47,
              109,
              97,
              112,
              47,
              119,
              104,
              105,
              108,
              101,
              47,
              119,
              104,
              105,
              108,
              101,
              95,
              99,
              111,
              110,
              116,
              101,
              120,
              116
            ]
          },
          "T": {
            "type": 20
          },
          "is_constant": {
            "b": true
          },
          "parallel_iterations": {
            "i": "32"
          }
        },
        "name": "Preprocessor/map/while/TensorArrayWrite/TensorArrayWriteV3/Enter",
        "op": "Enter"
      },
      {
        "input": [
          "Preprocessor/map/TensorArray"
        ],
        "attr": {
          "T": {
            "type": 20
          },
          "is_constant": {
            "b": true
          },
          "parallel_iterations": {
            "i": "32"
          },
          "frame_name": {
            "s": [
              80,
              114,
              101,
              112,
              114,
              111,
              99,
              101,
              115,
              115,
              111,
              114,
              47,
              109,
              97,
              112,
              47,
              119,
              104,
              105,
              108,
              101,
              47,
              119,
              104,
              105,
              108,
              101,
              95,
              99,
              111,
              110,
              116,
              101,
              120,
              116
            ]
          }
        },
        "name": "Preprocessor/map/while/TensorArrayReadV3/Enter",
        "op": "Enter"
      },
      {
        "input": [
          "Preprocessor/map/TensorArray",
          "Preprocessor/map/TensorArrayUnstack/range",
          "ToFloat",
          "Preprocessor/map/TensorArray:1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3",
        "op": "TensorArrayScatterV3"
      },
      {
        "input": [
          "Preprocessor/map/while/Merge",
          "Preprocessor/map/while/Less/Enter"
        ],
        "attr": {
          "T": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/Less",
        "op": "Less"
      },
      {
        "input": [
          "Preprocessor/map/while/Enter_1",
          "Preprocessor/map/while/NextIteration_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "N": {
            "i": "2"
          }
        },
        "name": "Preprocessor/map/while/Merge_1",
        "op": "Merge"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayUnstack/TensorArrayScatter/TensorArrayScatterV3"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "is_constant": {
            "b": true
          },
          "parallel_iterations": {
            "i": "32"
          },
          "frame_name": {
            "s": [
              80,
              114,
              101,
              112,
              114,
              111,
              99,
              101,
              115,
              115,
              111,
              114,
              47,
              109,
              97,
              112,
              47,
              119,
              104,
              105,
              108,
              101,
              47,
              119,
              104,
              105,
              108,
              101,
              95,
              99,
              111,
              110,
              116,
              101,
              120,
              116
            ]
          }
        },
        "name": "Preprocessor/map/while/TensorArrayReadV3/Enter_1",
        "op": "Enter"
      },
      {
        "input": [
          "Preprocessor/map/while/Less"
        ],
        "attr": {},
        "name": "Preprocessor/map/while/LoopCond",
        "op": "LoopCond"
      },
      {
        "input": [
          "Preprocessor/map/while/Merge",
          "Preprocessor/map/while/LoopCond"
        ],
        "attr": {
          "T": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/Switch",
        "op": "Switch"
      },
      {
        "input": [
          "Preprocessor/map/while/Merge_1",
          "Preprocessor/map/while/LoopCond"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/Switch_1",
        "op": "Switch"
      },
      {
        "input": [
          "Preprocessor/map/while/Switch:1"
        ],
        "attr": {
          "T": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/Identity",
        "op": "Identity"
      },
      {
        "input": [
          "Preprocessor/map/while/Switch_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/Exit_1",
        "op": "Exit"
      },
      {
        "input": [
          "Preprocessor/map/while/TensorArrayReadV3/Enter",
          "Preprocessor/map/while/Identity",
          "Preprocessor/map/while/TensorArrayReadV3/Enter_1"
        ],
        "attr": {
          "dtype": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/TensorArrayReadV3",
        "op": "TensorArrayReadV3"
      },
      {
        "input": [
          "^Preprocessor/map/while/Identity"
        ],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                0
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/ResizeImage/ExpandDims/dim",
        "op": "Const"
      },
      {
        "input": [
          "^Preprocessor/map/while/Identity"
        ],
        "attr": {
          "dtype": {
            "type": 3
          },
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "2"
                  }
                ]
              }
            }
          }
        },
        "name": "Preprocessor/map/while/ResizeImage/stack",
        "op": "Const"
      },
      {
        "input": [
          "^Preprocessor/map/while/Identity"
        ],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [
                1
              ],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": []
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/add/y",
        "op": "Const"
      },
      {
        "input": [
          "Preprocessor/map/TensorArray_1",
          "Preprocessor/map/while/Exit_1"
        ],
        "attr": {},
        "name": "Preprocessor/map/TensorArrayStack/TensorArraySizeV3",
        "op": "TensorArraySizeV3"
      },
      {
        "input": [
          "Preprocessor/map/while/TensorArrayReadV3",
          "Preprocessor/map/while/ResizeImage/ExpandDims/dim"
        ],
        "attr": {
          "Tdim": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/ResizeImage/ExpandDims",
        "op": "ExpandDims"
      },
      {
        "input": [
          "Preprocessor/map/while/Identity",
          "Preprocessor/map/while/add/y"
        ],
        "attr": {
          "T": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/add",
        "op": "Add"
      },
      {
        "input": [
          "Preprocessor/map/TensorArrayStack/range/start",
          "Preprocessor/map/TensorArrayStack/TensorArraySizeV3",
          "Postprocessor/Tile/multiples/1"
        ],
        "attr": {
          "Tidx": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/TensorArrayStack/range",
        "op": "Range"
      },
      {
        "input": [
          "Preprocessor/map/while/ResizeImage/ExpandDims",
          "Preprocessor/map/while/ResizeImage/stack"
        ],
        "attr": {
          "align_corners": {
            "b": false
          },
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/ResizeImage/ResizeBilinear",
        "op": "ResizeBilinear"
      },
      {
        "input": [
          "Preprocessor/map/while/add"
        ],
        "attr": {
          "T": {
            "type": 3
          }
        },
        "name": "Preprocessor/map/while/NextIteration",
        "op": "NextIteration"
      },
      {
        "input": [
          "Preprocessor/map/TensorArray_1",
          "Preprocessor/map/TensorArrayStack/range",
          "Preprocessor/map/while/Exit_1"
        ],
        "attr": {
          "element_shape": {
            "shape": {
              "dim": [
                {
                  "size": "300"
                },
                {
                  "size": "300"
                },
                {
                  "size": "3"
                }
              ]
            }
          },
          "dtype": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/TensorArrayStack/TensorArrayGatherV3",
        "op": "TensorArrayGatherV3"
      },
      {
        "input": [
          "Preprocessor/map/while/ResizeImage/ResizeBilinear"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "squeeze_dims": {
            "list": {
              "s": [],
              "i": [
                "0"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "Preprocessor/map/while/ResizeImage/Squeeze",
        "op": "Squeeze"
      },
      {
        "input": [
          "Preprocessor/mul/x",
          "Preprocessor/map/TensorArrayStack/TensorArrayGatherV3"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/mul",
        "op": "Mul"
      },
      {
        "input": [
          "Preprocessor/map/while/TensorArrayWrite/TensorArrayWriteV3/Enter",
          "Preprocessor/map/while/Identity",
          "Preprocessor/map/while/ResizeImage/Squeeze",
          "Preprocessor/map/while/Switch_1:1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/TensorArrayWrite/TensorArrayWriteV3",
        "op": "TensorArrayWriteV3"
      },
      {
        "input": [
          "Preprocessor/mul",
          "Preprocessor/sub/y"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/sub",
        "op": "Sub"
      },
      {
        "input": [
          "Preprocessor/map/while/TensorArrayWrite/TensorArrayWriteV3"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Preprocessor/map/while/NextIteration_1",
        "op": "NextIteration"
      },
      {
        "input": [
          "Preprocessor/sub",
          "FeatureExtractor/MobilenetV2/Conv/Conv2D/merged_input"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv/project/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Conv2D/merged_input"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise_weights"
        ],
        "attr": {
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise_weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_2/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_2/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise_weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_4/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise_weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_4/ArithmeticOptimizer/AddOpsRewrite_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_5/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_5/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise_weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/project/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_7/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Conv2D/merged_input"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_7/ArithmeticOptimizer/AddOpsRewrite_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_8/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/project/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_8/ArithmeticOptimizer/AddOpsRewrite_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_9/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_9/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise_weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/project/Conv2D/merged_input"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_11/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise_weights"
        ],
        "attr": {
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/project/Conv2D/merged_input"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_11/ArithmeticOptimizer/AddOpsRewrite_add"
        ],
        "attr": {
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_12/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_12/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6",
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6",
          "BoxPredictor_0/ClassPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "BoxPredictor_0/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/depthwise",
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_0/ClassPredictor_depthwise/depthwise",
          "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_0/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "Index": {
            "type": 3
          },
          "shrink_axis_mask": {
            "i": "1"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "begin_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          }
        },
        "name": "BoxPredictor_0/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/ClassPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_0/strided_slice",
          "BoxPredictor_0/stack/1",
          "Postprocessor/Tile/multiples/1",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          }
        },
        "name": "BoxPredictor_0/stack",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_0/strided_slice",
          "BoxPredictor_0/stack/1",
          "BoxPredictor_0/stack_1/2"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "BoxPredictor_0/stack_1",
        "op": "Pack"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "BoxPredictor_0/BoxEncodingPredictor_depthwise/Relu6",
          "BoxPredictor_0/BoxEncodingPredictor/weights"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_0/ClassPredictor_depthwise/Relu6",
          "BoxPredictor_0/ClassPredictor/weights"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_0/ClassPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_0/BoxEncodingPredictor/Conv2D",
          "BoxPredictor_0/BoxEncodingPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_0/BoxEncodingPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_0/ClassPredictor/Conv2D",
          "BoxPredictor_0/ClassPredictor/biases"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_0/ClassPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_0/BoxEncodingPredictor/BiasAdd",
          "BoxPredictor_0/stack"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_0/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_0/ClassPredictor/BiasAdd",
          "BoxPredictor_0/stack_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_0/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Conv2D/merged_input"
        ],
        "attr": {
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise_weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_14/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise_weights"
        ],
        "attr": {
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_14/ArithmeticOptimizer/AddOpsRewrite_add"
        ],
        "attr": {
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          },
          "N": {
            "i": "2"
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add",
        "op": "AddN"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/sub",
          "FeatureExtractor/MobilenetV2/expanded_conv_15/ArithmeticOptimizer/AddOpsRewrite_Leaf_1_add"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "_grappler:ArithmeticOptimizer:AddOpsRewriteStage": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/ArithmeticOptimizer/AddOpsRewrite_add",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_15/ArithmeticOptimizer/AddOpsRewrite_add",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/project/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/add_1",
          "FeatureExtractor/MobilenetV2/Conv_1/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/Conv_1/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv_1/Relu6",
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv_1/Relu6",
          "BoxPredictor_1/ClassPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv_1/Relu6"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "BoxPredictor_1/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/Conv_1/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Conv2D/merged_input"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/depthwise",
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_1/ClassPredictor_depthwise/depthwise",
          "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_1/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "begin_mask": {
            "i": "0"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          },
          "T": {
            "type": 3
          },
          "Index": {
            "type": 3
          },
          "shrink_axis_mask": {
            "i": "1"
          }
        },
        "name": "BoxPredictor_1/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/ClassPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_1/strided_slice",
          "BoxPredictor_1/stack/1",
          "Postprocessor/Tile/multiples/1",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          },
          "T": {
            "type": 3
          }
        },
        "name": "BoxPredictor_1/stack",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_1/strided_slice",
          "BoxPredictor_1/stack/1",
          "BoxPredictor_0/stack_1/2"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "BoxPredictor_1/stack_1",
        "op": "Pack"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_1/BoxEncodingPredictor_depthwise/Relu6",
          "BoxPredictor_1/BoxEncodingPredictor/weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_1/ClassPredictor_depthwise/Relu6",
          "BoxPredictor_1/ClassPredictor/weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "BoxPredictor_1/ClassPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise_weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "BoxPredictor_1/BoxEncodingPredictor/Conv2D",
          "BoxPredictor_1/BoxEncodingPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_1/BoxEncodingPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_1/ClassPredictor/Conv2D",
          "BoxPredictor_1/ClassPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_1/ClassPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "BoxPredictor_1/BoxEncodingPredictor/BiasAdd",
          "BoxPredictor_1/stack"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_1/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_1/ClassPredictor/BiasAdd",
          "BoxPredictor_1/stack_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_1/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Conv2D/merged_input"
        ],
        "attr": {
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6",
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6",
          "BoxPredictor_2/ClassPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "BoxPredictor_2/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/depthwise",
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_2/ClassPredictor_depthwise/depthwise",
          "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_2/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "Index": {
            "type": 3
          },
          "shrink_axis_mask": {
            "i": "1"
          },
          "begin_mask": {
            "i": "0"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          }
        },
        "name": "BoxPredictor_2/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_2/ClassPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_2/strided_slice",
          "BoxPredictor_2/stack/1",
          "Postprocessor/Tile/multiples/1",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          }
        },
        "name": "BoxPredictor_2/stack",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_2/strided_slice",
          "BoxPredictor_2/stack/1",
          "BoxPredictor_0/stack_1/2"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "BoxPredictor_2/stack_1",
        "op": "Pack"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_2/BoxEncodingPredictor_depthwise/Relu6",
          "BoxPredictor_2/BoxEncodingPredictor/weights"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_2/ClassPredictor_depthwise/Relu6",
          "BoxPredictor_2/ClassPredictor/weights"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_2/ClassPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "BoxPredictor_2/BoxEncodingPredictor/Conv2D",
          "BoxPredictor_2/BoxEncodingPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_2/BoxEncodingPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_2/ClassPredictor/Conv2D",
          "BoxPredictor_2/ClassPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_2/ClassPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "BoxPredictor_2/BoxEncodingPredictor/BiasAdd",
          "BoxPredictor_2/stack"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_2/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_2/ClassPredictor/BiasAdd",
          "BoxPredictor_2/stack_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_2/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Conv2D/merged_input"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6",
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6",
          "BoxPredictor_3/ClassPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "BoxPredictor_3/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/depthwise",
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_3/ClassPredictor_depthwise/depthwise",
          "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_3/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "shrink_axis_mask": {
            "i": "1"
          },
          "begin_mask": {
            "i": "0"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          },
          "Index": {
            "type": 3
          },
          "T": {
            "type": 3
          }
        },
        "name": "BoxPredictor_3/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/ClassPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_3/strided_slice",
          "BoxPredictor_3/stack/1",
          "Postprocessor/Tile/multiples/1",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          },
          "T": {
            "type": 3
          }
        },
        "name": "BoxPredictor_3/stack",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_3/strided_slice",
          "BoxPredictor_3/stack/1",
          "BoxPredictor_0/stack_1/2"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "BoxPredictor_3/stack_1",
        "op": "Pack"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_3/BoxEncodingPredictor_depthwise/Relu6",
          "BoxPredictor_3/BoxEncodingPredictor/weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_3/ClassPredictor_depthwise/Relu6",
          "BoxPredictor_3/ClassPredictor/weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "BoxPredictor_3/ClassPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "BoxPredictor_3/BoxEncodingPredictor/Conv2D",
          "BoxPredictor_3/BoxEncodingPredictor/biases"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_3/BoxEncodingPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_3/ClassPredictor/Conv2D",
          "BoxPredictor_3/ClassPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_3/ClassPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "BoxPredictor_3/BoxEncodingPredictor/BiasAdd",
          "BoxPredictor_3/stack"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_3/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_3/ClassPredictor/BiasAdd",
          "BoxPredictor_3/stack_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_3/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Conv2D/merged_input"
        ],
        "attr": {
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6",
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6",
          "BoxPredictor_4/ClassPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "BoxPredictor_4/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Conv2D/merged_input"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/depthwise",
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_4/ClassPredictor_depthwise/depthwise",
          "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_4/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "shrink_axis_mask": {
            "i": "1"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "begin_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          },
          "Index": {
            "type": 3
          },
          "T": {
            "type": 3
          }
        },
        "name": "BoxPredictor_4/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_4/strided_slice",
          "BoxPredictor_4/stack/1",
          "Postprocessor/Tile/multiples/1",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          }
        },
        "name": "BoxPredictor_4/stack",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_4/strided_slice",
          "BoxPredictor_4/stack/1",
          "BoxPredictor_0/stack_1/2"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "BoxPredictor_4/stack_1",
        "op": "Pack"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_4/BoxEncodingPredictor_depthwise/Relu6",
          "BoxPredictor_4/BoxEncodingPredictor/weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_4/ClassPredictor_depthwise/Relu6",
          "BoxPredictor_4/ClassPredictor/weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "BoxPredictor_4/ClassPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise_weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "2",
                "2",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "BoxPredictor_4/BoxEncodingPredictor/Conv2D",
          "BoxPredictor_4/BoxEncodingPredictor/biases"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/BoxEncodingPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_4/ClassPredictor/Conv2D",
          "BoxPredictor_4/ClassPredictor/biases"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_4/ClassPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/mul"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/mul_1",
        "op": "Mul"
      },
      {
        "input": [
          "BoxPredictor_4/BoxEncodingPredictor/BiasAdd",
          "BoxPredictor_4/stack"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_4/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_4/ClassPredictor/BiasAdd",
          "BoxPredictor_4/stack_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_4/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/Relu6",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Conv2D/merged_input"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/mul_1",
        "op": "Conv2D"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/mul_1",
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/add_1",
        "op": "Add"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6",
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6",
          "BoxPredictor_5/ClassPredictor_depthwise/depthwise_weights"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/depthwise",
        "op": "DepthwiseConv2dNative"
      },
      {
        "input": [
          "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Relu6"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "out_type": {
            "type": 3
          }
        },
        "name": "BoxPredictor_5/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/depthwise",
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          },
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_5/ClassPredictor_depthwise/depthwise",
          "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/gamma",
          "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/beta",
          "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/moving_mean",
          "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/moving_variance"
        ],
        "attr": {
          "epsilon": {
            "f": 0.0010000000474974513
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "is_training": {
            "b": false
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm",
        "op": "FusedBatchNorm"
      },
      {
        "input": [
          "BoxPredictor_5/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "new_axis_mask": {
            "i": "0"
          },
          "end_mask": {
            "i": "0"
          },
          "Index": {
            "type": 3
          },
          "T": {
            "type": 3
          },
          "shrink_axis_mask": {
            "i": "1"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "begin_mask": {
            "i": "0"
          }
        },
        "name": "BoxPredictor_5/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/FusedBatchNorm"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/ClassPredictor_depthwise/Relu6",
        "op": "Relu6"
      },
      {
        "input": [
          "BoxPredictor_5/strided_slice",
          "BoxPredictor_5/stack/1",
          "Postprocessor/Tile/multiples/1",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          },
          "T": {
            "type": 3
          }
        },
        "name": "BoxPredictor_5/stack",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_5/strided_slice",
          "BoxPredictor_5/stack/1",
          "BoxPredictor_0/stack_1/2"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "BoxPredictor_5/stack_1",
        "op": "Pack"
      },
      {
        "input": [
          "BoxPredictor_5/BoxEncodingPredictor_depthwise/Relu6",
          "BoxPredictor_5/BoxEncodingPredictor/weights"
        ],
        "attr": {
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          },
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_5/ClassPredictor_depthwise/Relu6",
          "BoxPredictor_5/ClassPredictor/weights"
        ],
        "attr": {
          "padding": {
            "s": [
              83,
              65,
              77,
              69
            ]
          },
          "dilations": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          },
          "strides": {
            "list": {
              "s": [],
              "i": [
                "1",
                "1",
                "1",
                "1"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          },
          "use_cudnn_on_gpu": {
            "b": true
          }
        },
        "name": "BoxPredictor_5/ClassPredictor/Conv2D",
        "op": "Conv2D"
      },
      {
        "input": [
          "BoxPredictor_5/BoxEncodingPredictor/Conv2D",
          "BoxPredictor_5/BoxEncodingPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_5/BoxEncodingPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_5/ClassPredictor/Conv2D",
          "BoxPredictor_5/ClassPredictor/biases"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "data_format": {
            "s": [
              78,
              72,
              87,
              67
            ]
          }
        },
        "name": "BoxPredictor_5/ClassPredictor/BiasAdd",
        "op": "BiasAdd"
      },
      {
        "input": [
          "BoxPredictor_5/BoxEncodingPredictor/BiasAdd",
          "BoxPredictor_5/stack"
        ],
        "attr": {
          "Tshape": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "BoxPredictor_5/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_5/ClassPredictor/BiasAdd",
          "BoxPredictor_5/stack_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "BoxPredictor_5/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "BoxPredictor_0/Reshape",
          "BoxPredictor_1/Reshape",
          "BoxPredictor_2/Reshape",
          "BoxPredictor_3/Reshape",
          "BoxPredictor_4/Reshape",
          "BoxPredictor_5/Reshape",
          "Postprocessor/Tile/multiples/1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "N": {
            "i": "6"
          },
          "Tidx": {
            "type": 3
          }
        },
        "name": "concat",
        "op": "ConcatV2"
      },
      {
        "input": [
          "BoxPredictor_0/Reshape_1",
          "BoxPredictor_1/Reshape_1",
          "BoxPredictor_2/Reshape_1",
          "BoxPredictor_3/Reshape_1",
          "BoxPredictor_4/Reshape_1",
          "BoxPredictor_5/Reshape_1",
          "Postprocessor/Tile/multiples/1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "N": {
            "i": "6"
          },
          "Tidx": {
            "type": 3
          }
        },
        "name": "concat_1",
        "op": "ConcatV2"
      },
      {
        "input": [
          "concat"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "squeeze_dims": {
            "list": {
              "s": [],
              "i": [
                "2"
              ],
              "f": [],
              "b": [],
              "type": [],
              "shape": [],
              "tensor": [],
              "func": []
            }
          }
        },
        "name": "Squeeze",
        "op": "Squeeze"
      },
      {
        "input": [
          "concat_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/convert_scores",
        "op": "Sigmoid"
      },
      {
        "input": [
          "Squeeze",
          "Postprocessor/Reshape_1/shape"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "Postprocessor/Reshape_1",
        "op": "Reshape"
      },
      {
        "input": [
          "Squeeze"
        ],
        "attr": {
          "out_type": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Shape",
        "op": "Shape"
      },
      {
        "input": [
          "Postprocessor/convert_scores",
          "Postprocessor/Slice/begin",
          "Postprocessor/Slice/size"
        ],
        "attr": {
          "Index": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Slice",
        "op": "Slice"
      },
      {
        "input": [
          "^Postprocessor/Reshape_1"
        ],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "2"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/Decode/transpose/sub_1",
        "op": "Const"
      },
      {
        "input": [
          "Postprocessor/Shape",
          "Postprocessor/strided_slice/stack",
          "Postprocessor/strided_slice/stack_1",
          "Postprocessor/strided_slice/stack_1"
        ],
        "attr": {
          "end_mask": {
            "i": "0"
          },
          "T": {
            "type": 3
          },
          "Index": {
            "type": 3
          },
          "shrink_axis_mask": {
            "i": "1"
          },
          "ellipsis_mask": {
            "i": "0"
          },
          "begin_mask": {
            "i": "0"
          },
          "new_axis_mask": {
            "i": "0"
          }
        },
        "name": "Postprocessor/strided_slice",
        "op": "StridedSlice"
      },
      {
        "input": [
          "Postprocessor/Reshape_1",
          "Postprocessor/Decode/transpose/sub_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tperm": {
            "type": 3
          }
        },
        "name": "Postprocessor/Decode/transpose",
        "op": "Transpose"
      },
      {
        "input": [
          "Postprocessor/strided_slice",
          "Postprocessor/Tile/multiples/1",
          "Postprocessor/Tile/multiples/1"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "Postprocessor/Tile/multiples",
        "op": "Pack"
      },
      {
        "input": [
          "Postprocessor/strided_slice",
          "MultipleGridAnchorGenerator/assert_equal/x",
          "BoxPredictor_0/stack/3"
        ],
        "attr": {
          "T": {
            "type": 3
          },
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "3"
          }
        },
        "name": "Postprocessor/stack",
        "op": "Pack"
      },
      {
        "input": [
          "Postprocessor/Decode/transpose"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "num": {
            "i": "4"
          },
          "axis": {
            "i": "0"
          }
        },
        "name": "Postprocessor/Decode/unstack",
        "op": "Unpack"
      },
      {
        "input": [
          "Postprocessor/ExpandDims",
          "Postprocessor/Tile/multiples"
        ],
        "attr": {
          "Tmultiples": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Tile",
        "op": "Tile"
      },
      {
        "input": [
          "Postprocessor/Decode/unstack",
          "ConstantFolding/Postprocessor/Decode/div_recip"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/div",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/unstack:2",
          "ConstantFolding/Postprocessor/Decode/div_2_recip"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/div_2",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/unstack:1",
          "ConstantFolding/Postprocessor/Decode/div_recip"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/div_1",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/unstack:3",
          "ConstantFolding/Postprocessor/Decode/div_2_recip"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/div_3",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Tile",
          "Postprocessor/Reshape_1/shape"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "Postprocessor/Reshape",
        "op": "Reshape"
      },
      {
        "input": [
          "Postprocessor/Decode/div_2"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/Exp_1",
        "op": "Exp"
      },
      {
        "input": [
          "Postprocessor/Decode/div_3"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/Exp",
        "op": "Exp"
      },
      {
        "input": [
          "^Postprocessor/Reshape"
        ],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "2"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/transpose/sub_1",
        "op": "Const"
      },
      {
        "input": [
          "Postprocessor/Reshape",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/transpose/sub_1"
        ],
        "attr": {
          "Tperm": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/transpose",
        "op": "Transpose"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/transpose"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "num": {
            "i": "4"
          },
          "axis": {
            "i": "0"
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack",
        "op": "Unpack"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack:2",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/sub_1",
        "op": "Sub"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack:3",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack:1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/sub",
        "op": "Sub"
      },
      {
        "input": [
          "Postprocessor/Decode/div",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/sub_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/mul_2",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/sub_1",
          "ConstantFolding/Postprocessor/Decode/div_4_recip"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/div",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/div_1",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/sub"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/mul_3",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/sub",
          "ConstantFolding/Postprocessor/Decode/div_4_recip"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/div_1",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/div"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/add",
        "op": "Add"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/div",
          "Postprocessor/Decode/Exp_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/div_4",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/unstack:1",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/div_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/add_1",
        "op": "Add"
      },
      {
        "input": [
          "Postprocessor/Decode/get_center_coordinates_and_sizes/div_1",
          "Postprocessor/Decode/Exp"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/div_5",
        "op": "Mul"
      },
      {
        "input": [
          "Postprocessor/Decode/mul_2",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/add"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/add",
        "op": "Add"
      },
      {
        "input": [
          "Postprocessor/Decode/mul_3",
          "Postprocessor/Decode/get_center_coordinates_and_sizes/add_1"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/add_1",
        "op": "Add"
      },
      {
        "input": [
          "Postprocessor/Decode/add",
          "Postprocessor/Decode/div_4"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/sub",
        "op": "Sub"
      },
      {
        "input": [
          "Postprocessor/Decode/add",
          "Postprocessor/Decode/div_4"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/add_2",
        "op": "Add"
      },
      {
        "input": [
          "Postprocessor/Decode/add_1",
          "Postprocessor/Decode/div_5"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/sub_1",
        "op": "Sub"
      },
      {
        "input": [
          "Postprocessor/Decode/add_1",
          "Postprocessor/Decode/div_5"
        ],
        "attr": {
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/add_3",
        "op": "Add"
      },
      {
        "input": [
          "Postprocessor/Decode/sub",
          "Postprocessor/Decode/sub_1",
          "Postprocessor/Decode/add_2",
          "Postprocessor/Decode/add_3"
        ],
        "attr": {
          "axis": {
            "i": "0"
          },
          "N": {
            "i": "4"
          },
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/Decode/stack",
        "op": "Pack"
      },
      {
        "input": [
          "^Postprocessor/Decode/stack"
        ],
        "attr": {
          "value": {
            "tensor": {
              "floatVal": [],
              "doubleVal": [],
              "intVal": [],
              "stringVal": [],
              "scomplexVal": [],
              "int64Val": [],
              "boolVal": [],
              "uint32Val": [],
              "uint64Val": [],
              "dtype": 3,
              "tensorShape": {
                "dim": [
                  {
                    "size": "2"
                  }
                ]
              }
            }
          },
          "dtype": {
            "type": 3
          }
        },
        "name": "Postprocessor/Decode/transpose_1/sub_1",
        "op": "Const"
      },
      {
        "input": [
          "Postprocessor/Decode/stack",
          "Postprocessor/Decode/transpose_1/sub_1"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tperm": {
            "type": 3
          }
        },
        "name": "Postprocessor/Decode/transpose_1",
        "op": "Transpose"
      },
      {
        "input": [
          "Postprocessor/Decode/transpose_1",
          "Postprocessor/stack"
        ],
        "attr": {
          "T": {
            "type": 1
          },
          "Tshape": {
            "type": 3
          }
        },
        "name": "Postprocessor/Reshape_2",
        "op": "Reshape"
      },
      {
        "input": [
          "Postprocessor/Reshape_2",
          "Postprocessor/ExpandDims_1/dim"
        ],
        "attr": {
          "Tdim": {
            "type": 3
          },
          "T": {
            "type": 1
          }
        },
        "name": "Postprocessor/ExpandDims_1",
        "op": "ExpandDims"
      }
    ],
    "library": {
      "function": [],
      "gradient": []
    },
    "versions": {
      "badConsumers": []
    }
  },
  "weightsManifest": [
    {
      "weights": [
        {
          "shape": [],
          "dtype": "float32",
          "name": "ConstantFolding/Postprocessor/Decode/div_recip"
        },
        {
          "shape": [
            1,
            1917,
            4
          ],
          "dtype": "float32",
          "name": "Postprocessor/ExpandDims"
        },
        {
          "shape": [],
          "dtype": "float32",
          "name": "ConstantFolding/Postprocessor/Decode/div_4_recip"
        },
        {
          "shape": [
            2
          ],
          "dtype": "int32",
          "name": "Postprocessor/Reshape_1/shape"
        },
        {
          "shape": [],
          "dtype": "float32",
          "name": "ConstantFolding/Postprocessor/Decode/div_2_recip"
        },
        {
          "shape": [
            3,
            3,
            576,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            576,
            12
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor/weights"
        },
        {
          "shape": [
            12
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/BoxEncodingPredictor/biases"
        },
        {
          "shape": [
            3,
            3,
            1280,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            1280,
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor/weights"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/BoxEncodingPredictor/biases"
        },
        {
          "shape": [
            3,
            3,
            512,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            512,
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor/weights"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/BoxEncodingPredictor/biases"
        },
        {
          "shape": [
            3,
            3,
            256,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            256,
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor/weights"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/BoxEncodingPredictor/biases"
        },
        {
          "shape": [
            3,
            3,
            256,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            256,
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor/weights"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/BoxEncodingPredictor/biases"
        },
        {
          "shape": [
            3,
            3,
            128,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            128,
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor/weights"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/BoxEncodingPredictor/biases"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "MultipleGridAnchorGenerator/assert_equal/x"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_0/stack/3"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "Postprocessor/ExpandDims_1/dim"
        },
        {
          "shape": [
            3,
            3,
            576,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            576,
            273
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor/weights"
        },
        {
          "shape": [
            273
          ],
          "dtype": "float32",
          "name": "BoxPredictor_0/ClassPredictor/biases"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_0/stack/1"
        },
        {
          "shape": [
            3,
            3,
            1280,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            1280,
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor/weights"
        },
        {
          "shape": [
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_1/ClassPredictor/biases"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_1/stack/1"
        },
        {
          "shape": [
            3,
            3,
            512,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            512,
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor/weights"
        },
        {
          "shape": [
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_2/ClassPredictor/biases"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_2/stack/1"
        },
        {
          "shape": [
            3,
            3,
            256,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            256,
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor/weights"
        },
        {
          "shape": [
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_3/ClassPredictor/biases"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_3/stack/1"
        },
        {
          "shape": [
            3,
            3,
            256,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            256,
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor/weights"
        },
        {
          "shape": [
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_4/ClassPredictor/biases"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_4/stack/1"
        },
        {
          "shape": [
            3,
            3,
            128,
            1
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor_depthwise/depthwise_weights"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/gamma"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/beta"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/moving_mean"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor_depthwise/BatchNorm/moving_variance"
        },
        {
          "shape": [
            1,
            1,
            128,
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor/weights"
        },
        {
          "shape": [
            546
          ],
          "dtype": "float32",
          "name": "BoxPredictor_5/ClassPredictor/biases"
        },
        {
          "shape": [
            160
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            160,
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            960,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/depthwise_weights"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            960,
            160
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_15/project/Conv2D/merged_input"
        },
        {
          "shape": [
            160
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            160,
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            960,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/depthwise_weights"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            960,
            160
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_14/project/Conv2D/merged_input"
        },
        {
          "shape": [
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            96,
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            576,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/depthwise_weights"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            576,
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_12/project/Conv2D/merged_input"
        },
        {
          "shape": [
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            96,
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            576,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/depthwise_weights"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            576,
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_11/project/Conv2D/merged_input"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            64,
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            384,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/depthwise_weights"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            384,
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_9/project/Conv2D/merged_input"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            64,
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            384,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/depthwise_weights"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            384,
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_8/project/Conv2D/merged_input"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            64,
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            384,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/depthwise_weights"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            384,
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_7/project/Conv2D/merged_input"
        },
        {
          "shape": [
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            32,
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            192,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/depthwise_weights"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            192,
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_5/project/Conv2D/merged_input"
        },
        {
          "shape": [
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            32,
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            192,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/depthwise_weights"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            192,
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_4/project/Conv2D/merged_input"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            24,
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            144,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/depthwise_weights"
        },
        {
          "shape": [
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            144,
            24
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_2/project/Conv2D/merged_input"
        },
        {
          "shape": [],
          "dtype": "float32",
          "name": "Preprocessor/mul/x"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "Preprocessor/map/TensorArrayStack/range/start"
        },
        {
          "shape": [],
          "dtype": "float32",
          "name": "Preprocessor/sub/y"
        },
        {
          "shape": [
            3,
            3,
            3,
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/Conv/Conv2D/merged_input"
        },
        {
          "shape": [
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/Conv/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            32,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/depthwise_weights"
        },
        {
          "shape": [
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            32,
            16
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv/project/Conv2D/merged_input"
        },
        {
          "shape": [
            16
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            16,
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            96,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/depthwise_weights"
        },
        {
          "shape": [
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            96,
            24
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/project/Conv2D/merged_input"
        },
        {
          "shape": [
            24
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_1/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            24,
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            144,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/depthwise_weights"
        },
        {
          "shape": [
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            144
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            144,
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/project/Conv2D/merged_input"
        },
        {
          "shape": [
            32
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_3/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            32,
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            192,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/depthwise_weights"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            192
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            192,
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/project/Conv2D/merged_input"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_6/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            64,
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            384,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/depthwise_weights"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            384
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            384,
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/project/Conv2D/merged_input"
        },
        {
          "shape": [
            96
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_10/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            96,
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            576,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/depthwise_weights"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            576
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            576,
            160
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/project/Conv2D/merged_input"
        },
        {
          "shape": [
            160
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_13/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            160,
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/Conv2D/merged_input"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/expand/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            960,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/depthwise_weights"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            960
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            960,
            320
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/project/Conv2D/merged_input"
        },
        {
          "shape": [
            320
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/expanded_conv_16/project/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            320,
            1280
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/Conv_1/Conv2D/merged_input"
        },
        {
          "shape": [
            1280
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/Conv_1/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            1280,
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/Conv2D/merged_input"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_2_1x1_256/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            256,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/depthwise_weights"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512_depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            256,
            512
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/Conv2D/merged_input"
        },
        {
          "shape": [
            512
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_2_3x3_s2_512/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            512,
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/Conv2D/merged_input"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_3_1x1_128/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            128,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/depthwise_weights"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256_depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            128,
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/Conv2D/merged_input"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_3_3x3_s2_256/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            256,
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/Conv2D/merged_input"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_4_1x1_128/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            128,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/depthwise_weights"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256_depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            128,
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/Conv2D/merged_input"
        },
        {
          "shape": [
            256
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_4_3x3_s2_256/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            256,
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/Conv2D/merged_input"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_1_Conv2d_5_1x1_64/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            3,
            3,
            64,
            1
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/depthwise_weights"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/mul"
        },
        {
          "shape": [
            64
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1,
            1,
            64,
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/Conv2D/merged_input"
        },
        {
          "shape": [
            128
          ],
          "dtype": "float32",
          "name": "FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128/BatchNorm/batchnorm/sub"
        },
        {
          "shape": [
            1
          ],
          "dtype": "int32",
          "name": "Postprocessor/strided_slice/stack"
        },
        {
          "shape": [
            1
          ],
          "dtype": "int32",
          "name": "Postprocessor/strided_slice/stack_1"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_5/stack/1"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "BoxPredictor_0/stack_1/2"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "Postprocessor/Tile/multiples/1"
        },
        {
          "shape": [
            3
          ],
          "dtype": "int32",
          "name": "Postprocessor/Slice/begin"
        },
        {
          "shape": [
            3
          ],
          "dtype": "int32",
          "name": "Postprocessor/Slice/size"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "Preprocessor/map/while/ResizeImage/ExpandDims/dim"
        },
        {
          "shape": [
            2
          ],
          "dtype": "int32",
          "name": "Preprocessor/map/while/ResizeImage/stack"
        },
        {
          "shape": [],
          "dtype": "int32",
          "name": "Preprocessor/map/while/add/y"
        },
        {
          "shape": [
            2
          ],
          "dtype": "int32",
          "name": "Postprocessor/Decode/transpose/sub_1"
        },
        {
          "shape": [
            2
          ],
          "dtype": "int32",
          "name": "Postprocessor/Decode/get_center_coordinates_and_sizes/transpose/sub_1"
        },
        {
          "shape": [
            2
          ],
          "dtype": "int32",
          "name": "Postprocessor/Decode/transpose_1/sub_1"
        }
      ],
      "paths": [
        "group1-shard1of5",
        "group1-shard2of5",
        "group1-shard3of5",
        "group1-shard4of5",
        "group1-shard5of5"
      ]
    }
  ]
}