Restormer-Plus for Real World Image Deraining: the Runner-up Solution to the GT-RAIN Challenge (CVPR 2023 UG2+ Track 3)
This is the Python code used to implement the Restormer-Plus method as described in the technical report:
This technical report presents our Restormer-Plus approach, which was submitted to the GT-RAIN Challenge (CVPR 2023 UG$^2$+ Track 3). Details regarding the challenge are available at http://cvpr2023.ug2challenge.org/track3.html. Our Restormer-Plus outperformed all other submitted solutions in terms of peak signal-to-noise ratio (PSNR). It consists mainly of four modules: the single image de-raining module, the median filtering module, the weighted averaging module, and the post-processing module. We named the single-image de-raining module Restormer-X, which is built on Restormer and performed on each rainy image. The median filtering module is employed as a median operator for the 300 rainy images associated with each scene. The weighted averaging module combines the median filtering results with that of Restormer-X to alleviate overfitting if we only use Restormer-X. Finally, the post-processing module is used to improve the brightness restoration. Together, these modules render Restormer-Plus to be one state-of-the-art solution to the GT-RAIN Challenge. Our code is available at https://github.com/ZJLAB-AMMI/Restormer-Plus.
The dataset can be found here.
- einops==0.3.0
- natsort==8.3.1
- numpy==1.21.5
- opencv_contrib_python==4.2.0.32
- Pillow==9.2.0
- piq==0.7.0
- skimage==0.0
- tabulate==0.8.10
- torch==1.12.1
- torchvision==0.13.1
Download the dataset from the link above and change the parameters in the train.py and test.py code to point to the appropriate directories (e.g., ./gt-rain/).
Download the pre-trained de-rain model from link.
Install all the required packages.
restormer-x:
-
training restormer baseline: set
model_version=baseand executepython /restormer_x/train.py. -
training restormer+: set
model_version=plusand executepython /restormer_x/train.py. -
evaluate and/or test: execute
python /restormer_x/test.py.
median: execute python /median/median_derain.py.
ensemble: execute python /ensemble/ensemble_derain.py.
post process: execute python /post_process/post_process_derain.py.
submit result: execute python repeat300.py.
If you find this code useful, please kindly cite
@article{zheng2023RestormerPlus,
title={Restormer-Plus for Real World Image Deraining: the Runner-up Solution to the GT-RAIN Challenge (CVPR 2023 UG2+ Track 3)},
author={Zheng, Chaochao, Wang, Luping and Liu, Bin},
journal={arXiv preprint arXiv:2305.05454},
year={2023}
}
Please only use the code and dataset for research purposes.
Chaochao Zheng
Zhejiang Lab, Research Center for Applied Mathematics and Machine Intelligence
zhengcc@zhejianglab.com
Luping Wang
Zhejiang Lab, Research Center for Applied Mathematics and Machine Intelligence
wangluping@zhejianglab.com