Neural Path Guiding with Distribution Factorization

TL;DR

This paper introduces a neural path guiding technique for Monte Carlo rendering that uses a novel distribution factorization approach, combining expressiveness and efficiency to improve rendering quality in complex scenes.

Contribution

We propose a simple neural representation that factorizes 2D distributions into two 1D PDFs, enhancing both speed and expressiveness for neural path guiding.

Findings

Outperforms existing methods in complex light transport scenes

Provides a balance of speed and accuracy in distribution modeling

Reduces variance in Monte Carlo integration through learned distributions

Abstract

In this paper, we present a neural path guiding method to aid with Monte Carlo (MC) integration in rendering. Existing neural methods utilize distribution representations that are either fast or expressive, but not both. We propose a simple, but effective, representation that is sufficiently expressive and reasonably fast. Specifically, we break down the 2D distribution over the directional domain into two 1D probability distribution functions (PDF). We propose to model each 1D PDF using a neural network that estimates the distribution at a set of discrete coordinates. The PDF at an arbitrary location can then be evaluated and sampled through interpolation. To train the network, we maximize the similarity of the learned and target distributions. To reduce the variance of the gradient during optimizations and estimate the normalization factor, we propose to cache the incoming radiance using an additional network. Through extensive experiments, we demonstrate that our approach is better than the existing methods, particularly in challenging scenes with complex light transport.