PSGLoss

Progressive Self-Guided Loss for Salient Object Detection

This repository contains the reference code for our TIP 2021 paper. The pdf can be found in this link

If you use any part of our code, or PSGLoss is useful for your research, please consider citing:

@ARTICLE{yang2021progressive,
  author={Yang, Sheng and Lin, Weisi and Lin, Guosheng and Jiang, Qiuping and Liu, Zichuan},
  journal={IEEE Transactions on Image Processing}, 
  title={Progressive Self-Guided Loss for Salient Object Detection}, 
  year={2021},
  volume={30},
  number={},
  pages={8426-8438},
  doi={10.1109/TIP.2021.3113794}}

model ckpt and results can be found in this link

Requirements

• Pytorch 0.4 !!!
• opencv-python

Train/Test

Train:

 CUDA_VISIBLE_DEVICES=0 python ystrain.py --batchsize 20 --lr 5e-5 --trainsize 352 --loss dicebce --randomflip --psgloss 

Test:

CUDA_VISIBLE_DEVICES=0 python test_ys.py --checkpointfile xxxx --batchsize 20 --lr 5e-5  --loss dicebce --testsize 352

Acknowledgments

Code and data prepration largely benefits from CPD