dazinovic / neural-rgbd-surface-reconstruction

Official implementation of the CVPR 2022 Paper "Neural RGB-D Surface Reconstruction"

Home Page:https://dazinovic.github.io/neural-rgbd-surface-reconstruction/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Result with real-world dataset is a mass, possible reasons

CherniGhassen opened this issue · comments

Hi,

When I run the model on my own dataset, I get a mass as you can see below.
Is it because of the translation/scaling factors that are applied? Or maybe because of the format of the poses that I use?
From what I have seen in previous questions, when the pose is wrong, people still get a sort of distorted mesh.

I would really appreciate your help!

mass

commented

Most likely the poses you used are incorrect. The project assumes OpenGL-style poses. I would suggest visualizing the rays that were generated with the provided poses to see if they make sense.