search

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 DiffAugment-biggan with the self-made dataset? #90

Closed Charlie-Wilson0211 closed 2 years ago

Charlie-Wilson0211 commented 2 years ago

If I want to train the few-shot dataset of 256*256 pixel resolution and less than 100 samples, and how to train DiffAugment-biggan with the few-shot dataset?

zsyzzsoft commented 2 years ago

This is not recommended as BigGAN does not perform as well as StyleGANs with very limited data. However, if you want to try out, you may refer to the original repo's README.

Charlie-Wilson0211 commented 2 years ago

Thanks for your help

arslanjavedghouri commented 1 year ago

@zsyzzsoft as you mentioned the link but in that code, there is no Differentiable Augmentation implemented to get good results with the limited dataset.