Plateau-reduced Differentiable Path Tracing

TL;DR

This paper introduces a method to improve differentiable rendering by smoothing the parameter space to avoid optimization plateaus, enabling better convergence in complex light transport scenarios.

Contribution

It proposes a convolution-based approach with Monte Carlo estimators to produce plateau-free gradients, enhancing optimization in differentiable rendering.

Findings

Improved convergence in complex light transport scenarios.

Reduced optimization error and runtime.

Effective extension to existing renderers.

Abstract

Current differentiable renderers provide light transport gradients with respect to arbitrary scene parameters. However, the mere existence of these gradients does not guarantee useful update steps in an optimization. Instead, inverse rendering might not converge due to inherent plateaus, i.e., regions of zero gradient, in the objective function. We propose to alleviate this by convolving the high-dimensional rendering function that maps scene parameters to images with an additional kernel that blurs the parameter space. We describe two Monte Carlo estimators to compute plateau-free gradients efficiently, i.e., with low variance, and show that these translate into net-gains in optimization error and runtime performance. Our approach is a straightforward extension to both black-box and differentiable renderers and enables optimization of problems with intricate light transport, such as caustics or global illumination, that existing differentiable renderers do not converge on.