The official PyTorch implementation of CVPR 2020 paper "Interactive Image Segmentation with First Click Attention".
- PyTorch>=0.4.1
- Opencv
- Scipy
- Matplotlib
Put the pretrained model into the created folder "pretrained_model" and the unzipped datasets into the created folder "dataset".
python evaluate.py --backbone [resnet,res2net] --dataset [GrabCut,Berkeley,DAVIS,VOC2012] (--sis)
python annotator.py --backbone [resnet,res2net] --input test.jpg --output test_mask.jpg (--sis)
([x,y,z] means choices. (x) means optional.)
- 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 )
- fcanet-resnet ( GoogleDrive | BaiduYun pwd: g989 )
- fcanet-res2net ( GoogleDrive | BaiduYun pwd: xzmx )
If you find this work or code is helpful in your research, please cite:
@inproceedings{fClick20CVPR,
title={Interactive Image Segmentation with First Click Attention},
author={Zheng Lin and Zhao Zhang and Lin-Zhuo Chen and Ming-Ming Cheng and Shao-Ping Lu},
booktitle={IEEE CVPR},
year={2020},
}
If you have any questions, feel free to contact me via: frazer.linzheng(at)gmail.com
.