S3IM: Stochastic Structural SIMilarity and Its Unreasonable Effectiveness for Neural Fields

TL;DR

This paper introduces S3IM, a novel nonlocal training loss for neural fields that leverages structural similarity across pixels, significantly enhancing rendering and reconstruction quality with minimal additional cost.

Contribution

The paper proposes S3IM, a new stochastic structural similarity loss that processes multiple pixels collectively, improving neural field training beyond traditional point-wise methods.

Findings

Test MSE drops over 90% for TensoRF and DVGO.

F-score increases by 198% for NeuS.

Robust performance with sparse, corrupted, and dynamic data.

Abstract

Recently, Neural Radiance Field (NeRF) has shown great success in rendering novel-view images of a given scene by learning an implicit representation with only posed RGB images. NeRF and relevant neural field methods (e.g., neural surface representation) typically optimize a point-wise loss and make point-wise predictions, where one data point corresponds to one pixel. Unfortunately, this line of research failed to use the collective supervision of distant pixels, although it is known that pixels in an image or scene can provide rich structural information. To the best of our knowledge, we are the first to design a nonlocal multiplex training paradigm for NeRF and relevant neural field methods via a novel Stochastic Structural SIMilarity (S3IM) loss that processes multiple data points as a whole set instead of process multiple inputs independently. Our extensive experiments demonstrate the unreasonable effectiveness of S3IM in improving NeRF and neural surface representation for nearly free. The improvements of quality metrics can be particularly significant for those relatively difficult tasks: e.g., the test MSE loss unexpectedly drops by more than 90% for TensoRF and DVGO over eight novel view synthesis tasks; a 198% F-score gain and a 64% Chamfer $L_{1}$ distance reduction for NeuS over eight surface reconstruction tasks. Moreover, S3IM is consistently robust even with sparse inputs, corrupted images, and dynamic scenes.