Closed tommy19970714 closed 2 years ago
This is very useful for editing features to produce animation effects,I can't wait to see this project support stylegan3.
We've actually already begun experimenting with various encoders for StyleGAN3, but training a proper encoder is a bit tricky. Feel free to take a look at my more detailed answer here: https://github.com/yuval-alaluf/restyle-encoder/issues/45#issuecomment-943503066
Interestingly, some initial experiments showed that current latent space editing techniques result in edits that are much more entangled than what you would get with SG2. Therefore, it seems like using SG3 for editing real images may lead to some undesirable results (using currently available techniques)
i trained one https://github.com/dayu1979/stylegan3encoder
Stumbled upon this old issue, thought to share the great work of Alaluf and Patashnik et al. Third Time's the Charm? Image and Video Editing with StyleGAN3 which enable StyleGAN3 training of e4e (ReStyle-e4e with 1 iteration).
StyleGAN3 is now available. https://github.com/NVlabs/stylegan3
Do you have plan to support StyleGAN3?