zhengyuf / IMavatar

Official repository for CVPR 2022 paper: I M Avatar: Implicit Morphable Head Avatars from Videos

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Clarify the pipeline

longphamkhac opened this issue · comments

Tks for your brilliant work!!! However, I want to clarify several steps in the proposed method. Specifically, first of all, the proposed method extract the deformed points by utilizing FLAME, and then query the canonical points using correspondence seach, is that right ?

Hi,

the deformed points are not extracted from FLAME, they are instead sampled with ray marching, as in most implicit representation rendering methods. And yes the deformed points are then mapped to the canonical space using correspondence search.

Yufeng

feel free to re-open the issue