Release SimpleCopyPaste Mask R-CNN (open-mmlab#8104)

* Release SimpleCopyPaste Mask R-CNN * set Version: v2.25.0 * fix a link * Modify model accuracy.
HYH-M87 · Jun 1, 2022 · 44af55f · 44af55f
1 parent 3b24fdd
commit 44af55f
Show file tree

Hide file tree

Showing 2 changed files with 98 additions and 6 deletions.
diff --git a/configs/simple_copy_paste/README.md b/configs/simple_copy_paste/README.md
@@ -18,12 +18,12 @@ Building instance segmentation models that are data-efficient and can handle rar
 
 Standard Scale Jittering(SSJ) resizes and crops an image with a resize range of 0.8 to 1.25 of the original image size, and Simple Copy-Paste(SCP) selects a random subset of objects from one of the images and pastes them onto the other image.
 
-| Backbone | Training schedule | Augmentation | batch size | box AP | mask AP |                                                                           Config                                                                           |
-| :------: | :---------------: | :----------: | :--------: | :----: | :-----: | :--------------------------------------------------------------------------------------------------------------------------------------------------------: |
-|   R-50   |        90k        |     SSJ      |     64     |  43.3  |  39.2   |   [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco.py)    |
-|   R-50   |        90k        |   SSJ+SCP    |     64     |  43.9  |  39.2   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco.py)  |
-|   R-50   |       270k        |     SSJ      |     64     |  43.5  |  39.1   |   [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco.py)   |
-|   R-50   |       270k        |   SSJ+SCP    |     64     |  45.1  |  40.3   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco.py) |
+| Backbone | Training schedule | Augmentation | batch size | box AP | mask AP |                                                                           Config                                                                           |                                                                                                                                                                                                                               Download                                                                                                                                                                                                                               |
+| :------: | :---------------: | :----------: | :--------: | :----: | :-----: | :--------------------------------------------------------------------------------------------------------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: |
+|   R-50   |        90k        |     SSJ      |     64     |  43.3  |  39.0   |   [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco.py)    |           [model](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco_20220316_181409-f79c84c5.pth) \| [log](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco_20220316_181409.log.json)           |
+|   R-50   |        90k        |   SSJ+SCP    |     64     |  43.8  |  39.2   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco.py)  |   [model](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco_20220316_181307-6bc5726f.pth) \| [log](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco_20220316_181307.log.json)   |
+|   R-50   |       270k        |     SSJ      |     64     |  43.5  |  39.1   |   [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco.py)   |         [model](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco_20220324_182940-33a100c5.pth) \| [log](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco_20220324_182940.log.json)         |
+|   R-50   |       270k        |   SSJ+SCP    |     64     |  45.1  |  40.3   | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco_20220324_201229-80ee90b7.pth) \| [log](https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco_20220324_201229.log.json) |
 
 ## Citation
 

diff --git a/configs/simple_copy_paste/metafile.yml b/configs/simple_copy_paste/metafile.yml
@@ -0,0 +1,92 @@
+Collections:
+  - Name: SimpleCopyPaste
+    Metadata:
+      Training Data: COCO
+      Training Techniques:
+        - SGD with Momentum
+        - Weight Decay
+      Training Resources: 32x A100 GPUs
+      Architecture:
+        - Softmax
+        - RPN
+        - Convolution
+        - Dense Connections
+        - FPN
+        - ResNet
+        - RoIAlign
+    Paper:
+      URL: https://arxiv.org/abs/2012.07177
+      Title: "Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation"
+    README: configs/simple_copy_paste/README.md
+    Code:
+      URL: https://github.com/open-mmlab/mmdetection/blob/v2.25.0/mmdet/datasets/pipelines/transforms.py#L2762
+      Version: v2.25.0
+
+Models:
+  - Name: mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco
+    In Collection: SimpleCopyPaste
+    Config: configs/simplecopypaste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco.py
+    Metadata:
+      Training Memory (GB): 7.2
+      Iterations: 270000
+    Results:
+      - Task: Object Detection
+        Dataset: COCO
+        Metrics:
+          box AP: 43.5
+      - Task: Instance Segmentation
+        Dataset: COCO
+        Metrics:
+          mask AP: 39.1
+    Weights: https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_270k_coco_20220324_182940-33a100c5.pth
+
+  - Name: mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco
+    In Collection: SimpleCopyPaste
+    Config: configs/simplecopypaste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco.py
+    Metadata:
+      Training Memory (GB): 7.2
+      Iterations: 90000
+    Results:
+      - Task: Object Detection
+        Dataset: COCO
+        Metrics:
+          box AP: 43.3
+      - Task: Instance Segmentation
+        Dataset: COCO
+        Metrics:
+          mask AP: 39.0
+    Weights: https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_32x2_90k_coco_20220316_181409-f79c84c5.pth
+
+  - Name: mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco
+    In Collection: SimpleCopyPaste
+    Config: configs/simplecopypaste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco.py
+    Metadata:
+      Training Memory (GB): 7.2
+      Iterations: 270000
+    Results:
+      - Task: Object Detection
+        Dataset: COCO
+        Metrics:
+          box AP: 45.1
+      - Task: Instance Segmentation
+        Dataset: COCO
+        Metrics:
+          mask AP: 40.3
+    Weights: https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_270k_coco_20220324_201229-80ee90b7.pth
+
+  - Name: mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco
+    In Collection: SimpleCopyPaste
+    Config: configs/simplecopypaste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco.py
+    Metadata:
+      Training Memory (GB): 7.2
+      Iterations: 90000
+    Results:
+      - Task: Object Detection
+        Dataset: COCO
+        Metrics:
+          box AP: 43.8
+      - Task: Instance Segmentation
+        Dataset: COCO
+        Metrics:
+          mask AP: 39.2
+    Weights: https://download.openmmlab.com/mmdetection/v2.0/simple_copy_paste/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco/mask_rcnn_r50_fpn_syncbn-all_rpn-2conv_ssj_scp_32x2_90k_coco_20220316_181307-6bc5726f.pth