Distort-and-recover-tensorflow-py3

Code for "Distort-and-Recover: Color Enhancement with Deep Reinforcement Learning", CVPR18

Overview

A version for python3 and tensorflow 1.10.0

The origin code from the writer Jongchan.

• Get vgg.pyz from VGG in Tensorflow.
• Before training,prepare MIT5K train/test images in separate folders (train/raw/, train/target/, test/raw/, test/target/). And edit the path in main.py accordingly.

Run the model

Train the model:

CUDA_VISIBLE_DEVICES=3 python3 main.py --prefix test_run --data-dir="/home/vikiqiu/data/mit5k"