Home

Awesome

RL-Restore [project page][paper]

:triangular_flag_on_post: Support arbitrary input size. Aug 25<br/> :triangular_flag_on_post: Add Python3 compatibility. Aug 6<br/> :triangular_flag_on_post: Training code is ready! Jun 15

Overview

<p align="center"> <img src='imgs/real_world.png' width = "100%"> </p>

Prerequisite

Test

Train

Acknowledgement

The DQN algorithm is modified from DQN-tensorflow.

Citation

@inproceedings{yu2018crafting,
 author = {Yu, Ke and Dong, Chao and Lin, Liang and Loy, Chen Change},
 title = {Crafting a Toolchain for Image Restoration by Deep Reinforcement Learning},
 booktitle = {Proceedings of IEEE Conference on Computer Vision and Pattern Recognition},
 pages={2443--2452},
 year = {2018} 
}