search

River-Zhang / GTA

[NeurIPS 23] Official repository for NeurIPS 2023 paper "Global-correlated 3D-decoupling Transformer for Clothed Avatar Reconstruction"
https://river-zhang.github.io/GTA-projectpage/
102 stars 4 forks source link

About HGPIFu #16

Open Followmeczx opened 5 months ago

Followmeczx commented 5 months ago

When estimating the human body geometry, the query operation is performed in HGPIFuNet. image The first step is to project the sampled point set onto the image plane. But I found that the transforms parameter is None. image image So in xyz = self.projection(points, calibs, transforms), only the points are rotated and translated. Are all points in the world coordinate system? The projection operation only converts points from the world coordinate system to the camera coordinate system after rotation and translation, and does not project further to the image plane. Please give me some help.

River-Zhang commented 5 months ago

Actually, all points are projected into [-1, 1] space cubes to perform operations. It is not strictly camera coordinate but normalized device coordinates (NDC).

Followmeczx commented 5 months ago

Where is the code projected into [-1, 1] space cubes？