This package contains the Python implementation of "Semi-supervised Transfer Learning for Image Rain Removal", in CVPR 2019.

Download synthesized data from here, as supervised training data. Put input images in './data/rainy_image_dataset/input' and ground truth images in './data/rainy_image_dataset/label'. Run /data/generate.m to generate HDF files as training data.

python training.py

python testing.py

Please cite this paper if you use the code:

@InProceedings{Wei_2019_CVPR,
author = {Wei, Wei and Meng, Deyu and Zhao, Qian and Xu, Zongben and Wu, Ying},
title = {Semi-Supervised Transfer Learning for Image Rain Removal},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition},
year = {2019}
}

The prior work using MoG for video deraining can be found here:

@InProceedings{wei2017should,
  title={Should we encode rain streaks in video as deterministic or stochastic?},
  author={Wei, Wei and Yi, Lixuan and Xie, Qi and Zhao, Qian and Meng, Deyu and Xu, Zongben},
  booktitle={Proceedings of the IEEE International Conference on Computer Vision},
  year={2017}
}

We use Deep Detail Network as our baseline. Thanks for sharing the code!

1. You are welcomed to add more real data.
2. You are welcomed to try more recent derain network as baseline.