Abstract

This work presents Neural Visibility Field (NVF), a novel uncertainty quantification method for Neural Radiance Fields (NeRF) applied to active mapping.  Our key insight is that regions not visible in the training views lead to inherently unreliable color predictions by NeRF at this region, resulting in increased uncertainty in the synthesized views. To achieve this, we propose to use a Bayesian network to composite position-based field uncertainty into ray-based uncertainty in camera observations. Consequently, NVF naturally assigns higher uncertainty to unobserved regions, aiding robots to select the most informative next viewpoints. Extensive evaluations show that NVF excels not only in uncertainty quantification but also in scene reconstruction for active mapping, outperforming existing methods.

Method

Our key insight: predictions on regions not visible in the training views are unreliable. By explicitly modeling the visibility of arbitrary views and distilling that into an implicit model, we can accurately and efficiently estimate uncertainty. To achieve this, we model the volume rendering process as a Hybrid Bayesian Network to bridge position-based field visibility and ray-based observation uncertainty, allowing us to unify both types of uncertainty in a principled framework.

Qualitative Results - Uncertainty Estimation

Qualitative Results - Active Mapping