frazerlin / fcanet

FCANet

The official PyTorch implementation of CVPR 2020 paper "Interactive Image Segmentation with First Click Attention".

Requirement

• PyTorch>=0.4.1
• Opencv
• Scipy
• Matplotlib

Useage

Put the pretrained model into the folder "pretrained_model" and the unzipped datasets into the folder "dataset".

Evalution:

python evaluate.py --backbone [resnet,res2net] --dataset [GrabCut,Berkeley,DAVIS,VOC2012] (--sis) 

Demo with UI:

python annotator.py --backbone [resnet,res2net] --input test.jpg --output test_mask.jpg (--sis)

([x,y,z] means choices. (x) means optional.)

Datasets

• GrabCut ( GoogleDrive | BaiduYun pwd: 6qkj )
• Berkeley ( GoogleDrive | BaiduYun pwd: 6bfw )
• DAVIS ( GoogleDrive | BaiduYun pwd: u5vd )
  (The three datasets we used are downloaded from f-BRS)
• PASCAL ( official )

Pretrained models

• fcanet-resnet ( GoogleDrive | BaiduYun pwd: g989 )
• fcanet-res2net ( GoogleDrive | BaiduYun pwd: xzmx )

Other framework

We offer the code in constructed by MindSpore framework.

Citation

If you find this work or code is helpful in your research, please cite:

@inproceedings{lin2020fclick,
  title={Interactive image segmentation with first click attention},
  author={Lin, Zheng and Zhang, Zhao and Chen, Lin-Zhuo and Cheng, Ming-Ming and Lu, Shao-Ping},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={13339--13348},
  year={2020}
}

Contact

If you have any questions, feel free to contact me via: frazer.linzheng(at)gmail.com.
Welcome to visit the project page or my home page.

License

The source code is free for research and education use only. Any comercial use should get formal permission first.