Low-illumination image enhancement using a conditional generative adversarial network

This repository is a GAN model for Low-illumination image ehancement

Dependence

• tensorflow==1.8.0
• numpy==1.16.0
• pillow==6.2.2

Usage

1. Create ./log and ./result at first.
2. The training images saved in ./dataset/high and ./dataset/low.The testing iamges saved in ./dataset/test.
3. python3 train.py for training.
4. You can modify the parameters in train.py before training.
5. python3 test.py for testing.
6. The weight of GAN is saved in ./log and the result of testing is saved in ./result.

Citation

黄鐄,陶海军,王海峰.条件生成对抗网络的低照度图像增强方法[J].中国图象图形学报,2019,24(12):2149-2158.