search

TimoBolkart / voca

This codebase demonstrates how to synthesize realistic 3D character animations given an arbitrary speech signal and a static character mesh.
https://voca.is.tue.mpg.de/en
1.14k stars 271 forks source link

Interpertation of %M in the camera calibration files #88

Closed m-muaz closed 2 years ago

m-muaz commented 2 years ago

Hi, Can anyone help me in understanding what does %M and the three rows of numbers mean in the camera calibration files? Is it a matrix related to some camera related projection matrix?

Thanks!

(Sample image attached)

image

chaiyujin commented 2 years ago

I attempted to project 3D mesh onto image with given camera calibration. I found that M is rotation matrix and following XYZ are the translation of camera extrinsics (R, t).

m-muaz commented 2 years ago

@chaiyujin Did you try to generate some depth images from the 3D mesh? Also, which library you use for projecting the 3D mesh to an image? Thanks!

chaiyujin commented 2 years ago

@Muaz231 Didn't try that. The 3D meshes are already in world coordinates, the z component of each vertex can be treated as depth, but sparse. You can sythesize depth image by rendering 3D meshes onto image plane, with the original z component (value from .ply, without any matrix operation) as "color" value. I use numpy to compute matrices and OpenGL to render.

m-muaz commented 2 years ago

@chaiyujin Thanks!