Neural Importance Sampling of Many Lights

TL;DR

This paper introduces a neural importance sampling method for Monte Carlo rendering that efficiently handles many lights by predicting light distributions at shading points, improving rendering quality and speed.

Contribution

It presents a neural network-based approach combined with light hierarchies and residual learning to enhance importance sampling in complex scenes with numerous light sources.

Findings

Outperforms existing methods in complex scenes

Reduces variance and improves rendering efficiency

Accelerates training convergence with residual learning

Abstract

We propose a neural approach for estimating spatially varying light selection distributions to improve importance sampling in Monte Carlo rendering, particularly for complex scenes with many light sources. Our method uses a neural network to predict the light selection distribution at each shading point based on local information, trained by minimizing the KL-divergence between the learned and target distributions in an online manner. To efficiently manage hundreds or thousands of lights, we integrate our neural approach with light hierarchy techniques, where the network predicts cluster-level distributions and existing methods sample lights within clusters. Additionally, we introduce a residual learning strategy that leverages initial distributions from existing techniques, accelerating convergence during training. Our method achieves superior performance across diverse and challenging scenes.