CycleGAN

tensorflow 2.0;

Description

This implementation of CycleGAN is a combination and upgrade version of vanhuyz's tf1 version and tf2's official implementation. And this project can be utilized on Windows, because many other's implementation needs a pakage of Addons for instance normalization.

I have implemented the instance normalization.

The code is quite easy to use. The only thing you have to change is the data.py to adapt to your own dataset.

In addition, I believe this code can help you learn the usage of tensorflow 2! I am a cordial tf-boy!!! No hesitate to ask me any problem about TF2!!!

Main Dependencies

tensorflow 2.0
Open CV
Python 3.6.5

Note

In the future, I will update the demo code for wgan. And I will show the result of the demo. This cyclegan is leverged for generating foggy cityscapes for my research work. And train.py has a few problems, I will rewrite it soon and upload it.