Awesome

RAS

This code is for the paper "Reverse Attention for Salient Object Detection".pdf

Pytorch Version

A PyTorch version is available at here.

Citation

@inproceedings{chen2018eccv, 
  author={Chen, Shuhan and Tan, Xiuli and Wang, Ben and Hu, Xuelong}, 
  booktitle={European Conference on Computer Vision}, 
  title={Reverse Attention for Salient Object Detection}, 
  year={2018}
}

@article{chen2020tip, 
  author={Chen, Shuhan and Tan, Xiuli and Wang, Ben and Lu, Huchuan and Hu, Xuelong and Fu, Yun}, 
  journal={IEEE Transactions on Image Processing}, 
  title={Reverse Attention Based Residual Network for Salient Object Detection},
  volume={29},  
  pages={3763-3776},
  year={2020}
}

Installing

Install prerequisites for Caffe (http://caffe.berkeleyvision.org/installation.html#prequequisites).
Build DSS [1] with cuDNN v5.1 for acceleration. Supposing the root directory of DSS is $DSS.

USE_CUDNN := 1

Copy the folder RAS to $DSS/example/.

Training

Prepare training dataset and its corresponding data list.
Download the Pre-trained VGG model (VGG-16) and copy it to $DSS/example/ras.
Change the dataset path in $DSS/example/RAS/train.prototxt.
Run solve.py in shell (or you could use IDE like Eclipse).

cd $DSS/example/RAS/
python solve.py

Testing

Change the dataset path in $DSS/example/RAS-tutorial_save.py.
Run jupyter notebook RAS-tutorial_save.ipynb.

Evaluation

We use the code of [1] for evaluation.

Pre-trained RAS model

Pre-trained RAS model on MSRA-B: Baidu drive(h7qj) and Google drive. Note that this released model is newly trained and is slightly different from the one reported in our paper.

Saliency Map

ECCV 2018: The saliency maps on 7 datasets are available at Baidu drive(zin5) and Google drive. TIP 2020: The saliency maps on 6 datasets are available at Google drive.

Reference

[1] Hou, Q., Cheng, M.M., Hu, X., Borji, A., Tu, Z., Torr, P.: Deeply supervised salient object detection with short connections. In: CVPR. (2017) 5300–5309.