Van Huy
Van Huy
It said that `NotFoundError: input_sample.jpg`. So you need to provide a input image with name `input_sample.jpg`. Of course, you can change the name as you want.
GAN convergence is a very difficult problem in practice. At the moment, just look at TensorBoard and stop training whenever you "feel" it enough. For this task, maybe 150k steps...
Actually that one is the same as the original code. Ironically, the original code is different from the paper at that part. I was trying to imitate the original code...
Thank you for your information. I've just realized that my implementation of PatchGAN was incorrect and I'm working on it. Regards,
I've removed the wrong part of PatchGAN and now the results should be better. But it takes time to confirm.
Results on training set after 37 epochs. * apple2orange: * orange2apple:
I tried on horse2zebra dataset but the results are still blurry. Maybe it is more difficult to translate horse2zebra than orange2apple. Did you train horse2zebra with the original pytorch version?
I fixed another issue and retrained horse2zebra. The results are quite promising. A screenshot after 87 epochs:
@TX2012LH just 1 line but critical https://github.com/vanhuyz/CycleGAN-TensorFlow/commit/960d1b085459856213d653d0f4e5a1c42dc1f0cf. G and F of cycle loss were reversed :flushed: