search

znxlwm / UGATIT-pytorch

Official PyTorch implementation of U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation
MIT License
2.51k stars 476 forks source link

I publish a Pytorch version, welcom to follow. #60

Open shoutOutYangJie opened 3 years ago

shoutOutYangJie commented 3 years ago

Thank you. The code and training code can be found here. https://github.com/shoutOutYangJie/Morph-UGATIT

vinduon commented 3 years ago

Dear sir,

I am a little confused with 2 CAM loss functions (5) and (6) in the paper. I don't know why one function uses BCE for generator, one uses MSE for discriminator? Can someone explain it? Many thanks.

shoutOutYangJie commented 3 years ago

Dear sir,

I am a little confused with 2 CAM loss functions (5) and (6) in the paper. I don't know why one function uses BCE for generator, one uses MSE for discriminator? Can someone explain it? Many thanks.

The two CAM losses have a different effects. The CAM in Generator helps to generate more realistic image. The CAM in D helps to distinct the real image and fake image.

vinduon commented 3 years ago

Thank you for your response, sir! I also do not know that what is the main function of auxiliary classifier in Generator and Discriminator? Are they the same in both?