/mws

Code for paper Multi-source weak supervision for saliency detection

Primary LanguagePython

Code for the paper

Code for the paper in CVPR2019, 'Multi-source weak supervision for saliency detection' (download the pdf file)

Results

score/datasets ECSSD HKU-IS PASCALS SOD OMRON DUTS-test SED1 SED2
max$F_\beta$ .878 .856 .790 .799 .718 .767 .902 .849
MAE .096 .084 .134 .167 .114 .096 .081 .097

Download result maps: OneDrive / GoogleDrive

Usage

Test

  1. Environment: python2.7, pytorch'1.0'

  2. Download models and put in the current folder

  3. Run

python main.py \
--img_dir 'path/to/images(.jpg)' \
--gt_dir 'path/to/ground-truth(.png)'

Train

Please checkout the other branch of this repo

Citation

@inproceedings{zeng2019multi,
  title={Multi-source weak supervision for saliency detection},
  author={Zeng, Yu and Zhuge, Yunzhi and Lu, Huchuan and Zhang, Lihe and Qian, Mingyang and Yu, Yizhou},
  booktitle={IEEE Conference on Computer Vision and Pattern Recognition},
  year={2019}
}