Evidential Neural Radiance Fields

TL;DR

Evidential Neural Radiance Fields introduce a probabilistic method for NeRFs that accurately estimates both aleatoric and epistemic uncertainties in scene reconstruction, enhancing safety in critical applications.

Contribution

The paper presents a novel evidential approach that integrates uncertainty quantification directly into NeRF rendering without additional computational costs.

Findings

Achieves state-of-the-art scene reconstruction fidelity.

Provides reliable uncertainty estimates for both aleatoric and epistemic uncertainties.

Demonstrates effectiveness across three benchmark datasets.

Abstract

Understanding sources of uncertainty is fundamental to trustworthy three-dimensional scene modeling. While recent advances in neural radiance fields (NeRFs) achieve impressive accuracy in scene reconstruction and novel view synthesis, the lack of uncertainty estimation significantly limits their deployment in safety-critical settings. Existing uncertainty quantification methods for NeRFs fail to separately capture both aleatoric and epistemic uncertainties. Among those that do quantify one or the other, many of them either compromise rendering quality or incur significant computational overhead to obtain uncertainty estimates. To address these issues, we introduce Evidential Neural Radiance Fields, a probabilistic approach that seamlessly integrates with the NeRF rendering process, enabling direct quantification of both aleatoric and epistemic uncertainties from a single forward pass. We compare multiple uncertainty quantification methods on three standardized benchmarks, where our approach demonstrates state-of-the-art scene reconstruction fidelity and uncertainty estimation quality. Code is available at https://github.com/KerryDRX/EvidentialNeRF.