/ANRN

Primary LanguagePython

ANRN Tweet

Angle-Based Non-local Recurrent Network for Single Image Deraining

Implementation detail for our paper Angle-Based Non-local Recurrent Network for Single Image Deraining, this code also includes further resaerch beyound this paper.

Dataset

Download these datasets to code/datasets

Rain100H and Rain100L: Google Drive or Baidu YUN

If you have questions regarding the dataset (please contact us: wangzefan@bupt.edu.cn, czhu@bupt.edu.cn)

Envs

  • Pytorch 1.0
  • Python 3+
  • cuda 9.0+

Training

$ cd code/
# train network
$ sh train.sh
# test dateset and record test results
$ sh test.sh
# use ssim.py to calculate SSIM 
$ python ssim.py

Citation

Please cite this paper in your publications if it helps your research:

@inproceedings{wang2020angle,
  title={Angle-Based Non-local Recurrent Network for Single Image Deraining},
  author={Wang, Zefan and Zhu, Chuang and Liu, Jun and Lin, WenHui and Liu, YaTing and Li, Chunxu},
  booktitle={2020 IEEE 6th International Conference on Computer and Communications (ICCC)},
  pages={2261--2264},
  year={2020},
  organization={IEEE}
}

Please also cite the following paper if you use the dataset:

Yang W, Tan R T, Feng J, et al. Deep joint rain detection and removal from a single image[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 1357-1366.

Contact