kcheng1021 / GaussianPro

[ICML2024] Official code for GaussianPro: 3D Gaussian Splatting with Progressive Propagation
https://kcheng1021.github.io/gaussianpro.github.io/
MIT License
551 stars 36 forks source link

Question about Propagation with rendered normal #22

Open hugoycj opened 3 months ago

hugoycj commented 3 months ago

Thank you for your excellent work! I am a bit confused about the implementation of Propagation. It seems that the current implementation does not utilize the rendered normal as described in the paper, since no rendered normal is loaded for propagation and geom_consistency is set to false, indicating that the normal for propagation is randomly initialized if I understand this function correctly.

I am wondering if it is possible to load the rendered normal for propagation, or if there are any experiments or insights to share regarding the performance between randomly initializing and using the rendered normal for initialization.

Ericgone commented 3 months ago

hi, how do you preprocess your own data? Specifically, how did you get the 'Mask' folder?