Overview

Summary:

• Jointly estimates lighting, material properties, and geometry from multi-view images of glossy objects under static illumination.
• Uses spherical Gaussians to approximate the environment and specular BRDF.
• Used Signed Distance Function to learn geometry.

Poses and camera parameters: Obtained using Colmap.

Images: Used sample synthetic dataset(kitty).

Drawbacks: Requires background/forground masks. It cannot handle human reflections in the object of interest in the captured data.

Paper: PhySG

Results:: The below results are for a PhySG model trained on the kitty synthetic data.