pixel-nerf icon indicating copy to clipboard operation
pixel-nerf copied to clipboard

About the 'absolute camera pose' and the 'relative camera pose'

Open JiuTongBro opened this issue 3 years ago • 1 comments

Hi, thanks for your impressive work.

However, when I read your paper, it mentions that your model only needs 'relative camera pose'. I can't understand what's the difference between 'absolute camera pose' and 'relative camera pose'.

I suppose is it about the world coordinate and the camera coordinate? If it is, I can't understand how do you get the correct position of sampled points in the scene, without the uasge of camara poses in world coordinates, since the rendered scene is under the world coordinate. And I read your code of gen_rays() in src/util/util.py, it seems there's no difference between your usage of camera pose and NeRF.

Could you please explain it to me?

JiuTongBro avatar Jul 24 '21 13:07 JiuTongBro

Hello. Does the "pose.txt" in the SRN-cars dataset contain the rotation matrix from camera coordinates to world coordinates and the world coordinates of the camera center?

caiyongqi avatar Jan 01 '22 08:01 caiyongqi