ch10tang/fnm.pytorch

A PyTorch implementation of Unsupervised Face Normalization with Extreme Pose and Expression in the Wild

fnm.pytorch's Issues

about VGGFace2 Github

Thanks for your code. I meet a problem when I want to download the face expert model from VGGFace2 Github but Github says this is not the web page you are looking for.
How can I download the face expert model?
Thanks!

about the generation inference

Missing key(s) in state_dict: "module.fc.filter.weight", "module.fc.filter.bias", "module.fc2.bias".
Unexpected key(s) in state_dict: "module.fc.weight", "module.fc.bias".
size mismatch for module.fc2.weight: copying a param with shape torch.Size([80013, 256]) from checkpoint, the shape in current model is torch.Size([79077, 256]).

I have try many checkpoints， It seems your model doesn't support the checkpoint loading in the generation？

请问这是官方团队的实现吗？

The results is bad

Can you tell me How can I train my network?
I try to train the metwork but the results is bad. The generated pictures are fuzzy even there is noise on the picture.
So, I want to know whether there are some specific ways to train the network.

datasets

你好，能否提供部分multi-pie数据集？

about the size

Can you offer training list and more detail of your model

Thanks for you offer this code, I read your code and want to train it，Did you clean the CASIA-WebFace or use all the data in it.
Looking forward to your reply