How to realize 3d controllable editing in your demo?

jgkwak95 / SURF-GAN

[ECCV 2022] Official Pytorch implementation of "Injecting 3D Perception of Controllable NeRF-GAN into StyleGAN for Editable Portrait Image Synthesis"

MIT License

126 stars 14 forks source link

How to realize 3d controllable editing in your demo? #2

Closed xuanyuzhang21 closed 2 years ago

xuanyuzhang21 commented 2 years ago

Thanks for your great work. Could you tell me how to realize 3d controllable face editing via the code in your project? Thanks a lot.

jgkwak95 commented 2 years ago

Hi! I'm not sure which model you referred to: SURF-GAN or 3D-controllable StyleGAN.

If the former, see 'discover_semantics.py' and check which att belongs to which layer & dimension. After that, run 'render_video_semantics.py' by giving target LiDj to control the attribute. In the case of 3D-controllable StyleGAN, the codes are not released. I'm not sure if this will be published or not, but if it does, it will be published in other repo. to prevent env conflicts or other issues. Thank you.