Physically Based Lighting for 3D Generative Models of Cars

Computing physically based lighting for car models created from a 3D aware generative model creates realistic imagery with accurate reflections.

Authors

author-image

By

 

""

 

Our method enables interactive physically based rendering (PBR) with 3D generative adversarial networks (GAN) under arbitrary illumination conditions in the form of environment maps and unconstrained camera navigation, including 360-degree rotations. We achieve this with three main contributions: a method to estimate poses of a dataset of car images, a generative pipeline for PBR, and an improved generative network architecture and training solution. We only require a dataset of images of the desired object class (cars, in our case) and a dataset of environment maps for training. Our method learns a disentangled representation of shading, enabling relighting with high-frequency reflections on shiny car bodies.