By Ruohao Guo, Dantong Niu, Liao Qu, Zhenbo Li

This is the official implementation of SOTR.

Note: The area of AP_S, AP_M and AP_L are calculated by segmentation mask without using bbox information.

• First install Detectron2 following the official guide: INSTALL.md.

• Then build SOTR with:

https://github.com/easton-cau/SOTR
cd SOTR
python setup.py build develop

• Then follow datasets/README.md to set up the datasets (e.g., MS-COCO).

• Evaluating

  • Download the trained models for COCO.

  • Run the following command

    python tools/train_net.py \
        --config-file configs/SOTR/R101.yaml \
        --eval-only \
        --num-gpus 4 \
        MODEL.WEIGHTS work_dir/SOTR_R101/SOTR_R101.pth
    

• Training

  • Run the following command

    python tools/train_net.py \
        --config-file configs/SOTR/R101.yaml \
        --num-gpus 4 \
    

Thanks Detectron2 and AdelaiDet contribution to the community!

The work is supported by National Key R&D Program of China (2020YFD0900204) and Key-Area Research and Development Program of Guangdong Province China (2020B0202010009).

If you want to improve the usability or any piece of advice, please feel free to contant directly ([email protected]).

Please consider citing our paper in your publications if the project helps your research. BibTeX reference is as follow.

@inproceedings{guo2021sotr,
  title={SOTR: Segmenting Objects with Transformers},
  author={Guo, Ruohao and Niu, Dantong and Qu, Liao and Li, Zhenbo},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={7157--7166},
  year={2021}
}