search

KelestZ / CoRF

Controllable Radiance Fields for Dynamic Face Synthesis, 3DV 2022.
20 stars 1 forks source link

face reenactment on real images #1

Open yataoz opened 1 year ago

yataoz commented 1 year ago

Thanks for the great work!

I was wondering how you preserved the subject identity/appearance after face reenactment on real images? Did you use some GAN inversion approach to compute the latent W space?

sunshineatnoon commented 1 year ago

Same question here

KelestZ commented 1 year ago

Thanks for the great work!

I was wondering how you preserved the subject identity/appearance after face reenactment on real images? Did you use some GAN inversion approach to compute the latent W space?

Thanks for your interest and sorry for the late reply. We didn't try real-face reenactment on real images in the paper. The source images are synthetic faces. However, it's totally doable for real-face animation, i.e., one can definitely learn an encoder to invert the face image to the latent space.

KelestZ commented 1 year ago

Same question here

Thanks for your interest and sorry for the late reply. We didn't try real-face reenactment on real images in the paper. The source images are synthetic faces. However, it's totally doable for real-face animation, i.e., one can definitely learn an encoder to invert the face image to the latent space.