Skip to content

bhushan23/Illumination-Estimation-With-Residue-Networks

Repository files navigation

SC-Net

Shading Correcting Network

Measure issue with reconstructin faces using Shading, Albedo, Normal and Spherical Harmonics is little variations in normal and spherical harmonics causing different illumination.

Estimating Normal and Spherical Harmonics with slight error causes different illumination and expected. In order to overcome this problem, we would like to using correcting networks

Shading Correcting Network

1_Shading_Correcting works on priciple built on top of tradition shading generation using normal and spherical harmonics We generate Shading and then rectify it using latent ligthing generated from Normal, Albedo and Image features.

Latent Shading Generation

2_Latent_Shading_Gen works on principle to avoid generation of Spherical Harmonics and instead generate shading directly using Normal, Albedo and Image features. And then use this latent lighting along with Normal to generate shading

About

Illumination model based on Normal and Spherical Harmonics is not precise. Here, we introduce new illumination model based on residue of shading

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors