How to train StyleGAN2 with labeled dataset

mit-han-lab / data-efficient-gans

[NeurIPS 2020] Differentiable Augmentation for Data-Efficient GAN Training

https://arxiv.org/abs/2006.10738

BSD 2-Clause "Simplified" License

1.27k stars 175 forks source link

How to train StyleGAN2 with labeled dataset #58

Open TccccD opened 3 years ago

TccccD commented 3 years ago

Hello, I found that compared with the native stylegan2, your code specifically removes the label. I want to ask you whether you have tried to add the label for training, how is the effect?

TccccD commented 3 years ago

@zsyzzsoft

zsyzzsoft commented 3 years ago

I did try conditional generation before, and you will be able to do it if you make some changes in loss.py (e.g. something like real_scores = reduce_sum(D(reals) * one_hot(label)). Or you may try the PyTorch version to run conditional generation.