search

taconite / IntrinsicAvatar

[CVPR 2024] IntrinsicAvatar: Physically Based Inverse Rendering of Dynamic Humans from Monocular Videos via Explicit Ray Tracing
https://neuralbodies.github.io/IntrinsicAvatar/
MIT License
75 stars 5 forks source link

How to get the rendering results of normal from 'Geometry Grid and MLP' #5

Open JiatengLiu opened 3 months ago

JiatengLiu commented 3 months ago

Hello! I want to see the results of normal extracted from Geometry Grid and MLP Module, is there any scripts to finish this? Waiting for your reply!

taconite commented 2 months ago

Hi, Jiateng

I am not sure what you mean by "extracting normals". The normal results presented in the paper are all rendered via alpha composition from the gradients of the SDF field.

If you mean mesh extraction, you can refer to this function

best

JiatengLiu commented 2 months ago

Sorry for not being clear, what I mean is how to get the result of normal image rendered by the Geometry Grid and MLP module

taconite commented 2 months ago

The normal maps are rendered here: https://github.com/taconite/IntrinsicAvatar/blob/d243d20c4bfa5df10e6613c986a32bd76e7d1d3e/models/intrinsic_avatar.py#L1249-L1285

The corresponding functions are: https://github.com/taconite/IntrinsicAvatar/blob/d243d20c4bfa5df10e6613c986a32bd76e7d1d3e/models/volrend.py#L638 https://github.com/taconite/IntrinsicAvatar/blob/d243d20c4bfa5df10e6613c986a32bd76e7d1d3e/models/volrend.py#L810