Yijunmaverick/CartoonGAN-Test-Pytorch-Torch

can you share hyper-parameter ?

taki0112 opened this issue · 5 comments

taki0112 commented

Hi
I'm interested your paper
So I'm reproducing the CartoonGAN

However, I don't know hyper-paramter in detail.. like weight initialization.. (xavier ? gaussian?)
I want to know hyper-parameter information

  1. Optimizer (if Adam, what beta1, beta2` ?)
  2. Weight initialization
  3. Use l2 regularization (weight decay) ?
  4. Data augmentation
  5. Normalization function in discriminator (batch norm? layer norm?)
  6. Loss function (general gan loss ? or lsgan loss like cyclegan ?)

and can you share the animation dataset used in your paper?

Thank you !

Yijunmaverick commented

@taki0112 Sorry, this is not my work. You can ask authors for details.

taki0112 commented

@Yijunmaverick
I have a question about vgg's conv4_4 feature map.
Is this a feature map after relu? Or is it a feature map before relu?

FlyingGoblin commented

@Yijunmaverick before relu.
I am sorry I didn't clean up my code because of the graduation. I will try my best to upload my code in this month.
For the photo, all the images in our paper are from CycleGAN: https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix (XXX2photo).
Because all the cartoon images are from films, I don't think I have the right to upload them. I don't know if there will be any copyright problem， I am sorry.

Yijunmaverick commented

@taki0112 Check the author's reply above