Code for initializing NeRF's geometry as a sphere

Kai-46 / nerfplusplus

improves over nerf in 360 capture of unbounded scenes

BSD 2-Clause "Simplified" License

911 stars 101 forks source link

Code for initializing NeRF's geometry as a sphere #26

Open yd-yin opened 3 years ago

yd-yin commented 3 years ago

Hi! Thanks for the elaborate analysis and informative experiments!

Noticed that you conducted an experiment where the geometry (opacity field) of NeRF is fixed (or initialized). Could you please also offer the code of that experiment?

Thanks in advance!

Kai-46 commented 3 years ago

Sorry, I'm not sure if I can find the scripts. But training a opacity field to a unit sphere seems very straightforward? You can simply train the MLP using a loss to ask it to predict 1 on the sphere (x^2+y^2+z^2=1), and zero elsewhere.

joeyw1030 commented 1 year ago

你好！感谢您的精心分析和翔实的实验！

注意到您进行了一个实验，其中 NeRF 的几何形状（不透明度场）是固定的（或初始化的）。您能否也提供该实验的代码？

提前致谢！

Hello, have you solved this problem? How to initialize a 3D model into a sphere?

Rlee719 commented 1 year ago

Refer to the geometric network initialization method from this paper:

SAL: Sign Agnostic Learning of Shapes from Raw Data