/Self-ShadowGAN

Official implementation of "Learning to Remove Shadows from a Single Image"(Self-ShadowGAN)

Primary LanguagePython

[Self-ShadowGAN] Learning to Remove Shadows from a Single Image

This is the PyTorch implementation of the paper "Learning to Remove Shadows from a Single Image" published on International Journal of Computer Vision (IJCV).

Requirements

We training our model using

  • python 3.7 (or newer version)
  • pytorch 1.4 (or newer version)

Training & Inference

We put several examples (shadow image/mask) in ./images folder. You can use following code to run our method and get shadow removal result in ./output folder:

python main.py --gpu gpu_idx --idx image_idx (e.g. python main.py --gpu 0 --idx 1)

Or, you can put your own shadow images/masks into ./images folder and modified thier names.

Notes

  • If you use the code, please cite our paper.
@article{jiang2023learning,
  title={Learning to Remove Shadows from a Single Image},
  author={Jiang, Hao and Zhang, Qing and Nie, Yongwei and Zhu, Lei and Zheng, Wei-Shi},
  journal={International Journal of Computer Vision},
  volume={131},
  number={9},
  pages={2471--2488},
  year={2023},
  publisher={Springer}
}