README.md

Preparing Dataset

Preparing Training Dataset

We transfer our training pseudo labels to tsv format for faster dataloading. Please set the environment variable

export TRAIN_DATASETS=path_to_your_tsv_directory

The structure and file names of the tsv directory should follow

$TRAIN_DATASETS/
  SAM-1.tsv
  SAM-1.lineidx
  SAM-2.tsv
  SAM-2.lineidx
  ......
  SAM-N.tsv
  SAM-N.lineidx

Please refer to Semantic-SAM for a detailed script to transform json format data to tsv format data

Preparing Evaluation Dataset (Whole-Image-Segmentation)

By default we support 7 evaluation datasets for whole-image-segmentation evaluation: SA-1B, COCO, LVIS, ADE20K, EntitySeg, PartImagenet, and PACO. Please set the root directory environment variable first

export DETECTRON2_DATASETS=path_to_your_root_evaluation_directory

and change the test dataset name in config file

cfg.DATASETS.TEST: ("unsam_{sa1b, ade20k, entity, paco, partimagenet, coco, lvis}_val")

The structure and file names of the tsv directory should follow

$DETECTRON2_DATASETS/
  sa1b/
    images/
    annotations/
      sa1b_val.json
  ade/
    images/
    annotations/
      ade_val.json
  entity/
    images/
    annotations/
      entityseg_val.json
  paco/
    images/
    annotations/
      paco_val.json
  partimagenet/
    images/
    annotations/
      partimagenet_val.json
  coco/
    val2017/
    annotations/
      instances_val2017.json
  lvis/
    images/
    annotations/
      lvis_v1_val.json

Since ade20k, entity, partimagenet don't have standard image id format, please name each image as {image_id}.jpg. You can adjust them in whole_image_segmentation/data/build.py/get_test_detection_datasets

Preparing Evaluation Dataset (Promptable-Segmentation)

We support COCO evaluation for promptable segmentation. Please set the root directory environment variable first

export DETECTRON2_DATASETS=path_to_your_root_evaluation_directory

and refer to MaskDINO to prepare files under the evaluation directory