Stylegan2-ada
johndpope opened this issue · 3 comments
Was any efforts made to upgrade to stylegan2-ada?
Issue Label Bot is not confident enough to auto-label this issue. See dashboard for more details.
No, not as far as I am concerned.
StyleGAN2-ADA introduces an adaptive strategy for non-leaky augmentation during training with limited data. It does not innovate on the projection process.
Moreover, there are different configs, so I would leave it to the user to adapt the StyleGAN2-ADA projection code for their own config. It is not a big issue. In my case, it only required removing an AssertionError as mentioned in NVlabs/stylegan2-ada#26
Okay, one thing that the ADA repository makes easier for the user is that it saves the latent vector to out/dlatents.npz.
As for the W+ projection, it was implemented in https://github.com/pbaylies/stylegan2-ada with this commit pbaylies/stylegan2-ada@5ed6189#diff-76ff8c644e89292874e4cfdf91500e696f51aa28e2912b5f6b41f92955384d1c