• python 3.6.13
• pytorch 1.10.0
• numpy==1.19.2
• opencv_python_headless==4.5.4.60

1. You can obtain RDD dataset from here
2. Modify the config.yaml to set your parameters
3. Use extractBackground.py to construct the ground-truth background
4. Training CBENet

python Train_CBENet.py ./configs/model=CBENet/config.yaml

5. Training BGShadowNet

python Train_BGShadowNet.py ./configs/model=BGShadowNet/config.yaml

6. Pretrained model

Download the pretrained model trained on RDD dataset.

python test.py

  @InProceedings{Zhang_2023_CVPR,
  author  = {Ling Zhang,Yinghao He,Qing Zhang,Zheng Liu,Xiaolong Zhang,Chunxia Xiao},
  title   = {Document Image Shadow Removal Guided by Color-Aware Background.},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month   = {June},
  year   = {2023}}