Differentiable Shadow Mapping for Efficient Inverse Graphics

TL;DR

This paper introduces a method for efficiently generating differentiable shadows in rendering, significantly improving convergence and speed in inverse graphics tasks by combining shadow mapping with differentiable rasterization.

Contribution

It presents a novel approach that integrates pre-filtered shadow mapping with differentiable rasterizers to produce fast, accurate differentiable shadow maps for inverse graphics.

Findings

Differentiable shadow maps are much faster than traditional light transport simulation.

The method achieves similar accuracy with significantly reduced computation time.

Without shadows, differentiable rasterization often fails to converge.

Abstract

We show how shadows can be efficiently generated in differentiable rendering of triangle meshes. Our central observation is that pre-filtered shadow mapping, a technique for approximating shadows based on rendering from the perspective of a light, can be combined with existing differentiable rasterizers to yield differentiable visibility information. We demonstrate at several inverse graphics problems that differentiable shadow maps are orders of magnitude faster than differentiable light transport simulation with similar accuracy -- while differentiable rasterization without shadows often fails to converge.