xg-chu / GAGAvatar

[NeurIPS 2024] Generalizable and Animatable Gaussian Head Avatar
MIT License
202 stars 22 forks source link

question #15

Open xz0305 opened 1 day ago

xz0305 commented 1 day ago

Hello, after reading your paper, I'm having some difficulty understanding this part: 'the rotation of the head is modeled through changes in camera pose to ensure that the head itself is relatively stationary.' Why is the head movement described using camera pose instead of directly modifying the head's movement? What is the reason for keeping the head stationary?