Run in Torch7 platform.
I'm very appreciative of your attention. Because the trained models may be produce some different effects in varying platforms and computers, I strongly suggest that you can adopt our trained model when restoring real world haze images. If you want to repeate our works on other platforms (e.g. Pytorch), please referrencing https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix and finishing other changes. This demo is only for learning and communication.
Testing: th test.lua
Training: th train.lua
#Trained model is available on Baidu Pan
链接：https://pan.baidu.com/s/1Xk6eLyNG7QSrvc9-P6r4RQ 密码：uagt
#Trained model on the dataset with random noise is also available on Baidu Pan