Fast and Accurate Neural Rendering Using Semi-Gradients

TL;DR

This paper introduces a neural rendering framework that uses a novel unbiased and low-variance gradient-based loss function, resulting in faster training and more accurate global illumination rendering.

Contribution

It proposes a new objective function for neural rendering that reduces bias and variance in gradient estimates, improving training speed and accuracy.

Findings

Faster training convergence demonstrated in experiments.

More accurate rendering results compared to previous residual-based methods.

The method is simple to implement and theoretically sound.

Abstract

We propose a simple yet effective neural network-based framework for global illumination rendering. Recently, rendering techniques that learn neural radiance caches by minimizing the difference (i.e., residual) between the left and right sides of the rendering equation have been suggested. Due to their ease of implementation and the advantage of excluding path integral calculations, these techniques have been applied to various fields, such as free-viewpoint rendering, differentiable rendering, and real-time rendering. However, issues of slow training and occasionally darkened renders have been noted. We identify the cause of these issues as the bias and high variance present in the gradient estimates of the existing residual-based objective function. To address this, we introduce a new objective function that maintains the same global optimum as before but allows for unbiased and low-variance gradient estimates, enabling faster and more accurate training of neural networks. In conclusion, this method is simply implemented by ignoring the partial derivatives of the right-hand side, and theoretical and experimental analyses demonstrate the effectiveness of the proposed loss.