CycleGAN-TensorFlow icon indicating copy to clipboard operation
CycleGAN-TensorFlow copied to clipboard

identity mapping loss

Open iariav opened this issue 7 years ago • 5 comments

Hi, thank you for sharing this implementation. do you plan on adding support for the identity mapping loss as described in the original paper? thanks

iariav avatar May 23 '17 08:05 iariav

Yes, I will add identity mapping loss soon. But before that, I have to fix #18. Anyway, pull requests are always welcome! 😉

vanhuyz avatar May 23 '17 14:05 vanhuyz

Have you added the identity mapping loss yet? Besides, I found the training is slow. Do you have any idea to facilitate the training? What do you think of using WGAN-GP instead of LSGAN?

lizhuorong avatar Aug 09 '17 12:08 lizhuorong

@vanhuyz Good luck

Zxr1314 avatar Jul 26 '18 13:07 Zxr1314

@lizhuorong CycleWGAN should be a more robust method

Zxr1314 avatar Jul 26 '18 13:07 Zxr1314

i want to change the loss function by pixel wise loss , how i can do that please

hala3 avatar Jun 13 '19 09:06 hala3