about choosing latent dimension

mchong6 / GANsNRoses

Official PyTorch repo for GAN's N' Roses. Diverse im2im and vid2vid selfie to anime translation.

MIT License

1.16k stars 151 forks source link

about choosing latent dimension #23

Closed LeeDongYeun closed 3 years ago

LeeDongYeun commented 3 years ago

Hi, thanks for your great work!

I have a question about the latent dimension setting. Is there any specific reason to choose latent dimension to 8?

The base setting of stylegan's latent dimension is 512. And in just my opinion, dimension 8 is not enough to embed style. Could you explain why you set it up like that?

mchong6 commented 3 years ago

I followed the setting used in MUNIT. I haven't tried with higher dimension counts but potentially if the latent dimension is too high, the style code might encode "too much information" and may lead to content code being ignored.

LeeDongYeun commented 3 years ago

Thanks for reply!!! :)