bennyguo / instant-nsr-pl

Neural Surface reconstruction based on Instant-NGP. Efficient and customizable boilerplate for your research projects. Train NeuS in 10min!
MIT License
856 stars 84 forks source link

Math behind the color network. #117

Closed weihan1 closed 4 months ago

weihan1 commented 1 year ago

In the original NeuS paper, the input to the color network is direction encoding, feature, normal, sdf, and the positions stacked together (see below).

Screenshot 2023-09-21 at 6 55 49 PM

However, in the code, the input to the color network is only the direction encoding, feature, and normal, without the sdf or the positions. Why is that?

rgb = self.texture(feature, t_dirs, normal)