use conditional gan network to dehaze
the repository refer to
DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks
original paper use the network to deblur,in the repository use the network to dehaze
use utils/add_haze.py tool to genernate some haze image,and modify train.py ori_image path and haze_image path
python train.py
the model result save folder checkpoints/
Put the images you want to test into the folder test/
modify demo.py model path to lastest model
python demo.py