NeRF++: Analyzing and Improving Neural Radiance Fields

TL;DR

NeRF++ analyzes the limitations of neural radiance fields, particularly shape-radiance ambiguity, and proposes improvements to enhance view synthesis in large-scale, unbounded 3D scenes.

Contribution

The paper provides an analysis of radiance field ambiguities and introduces a new parametrization to improve NeRF's performance in unbounded scenes.

Findings

Addresses shape-radiance ambiguity in NeRF

Improves view synthesis fidelity for large-scale scenes

Provides a new parametrization method for unbounded scenes

Abstract

Neural Radiance Fields (NeRF) achieve impressive view synthesis results for a variety of capture settings, including 360 capture of bounded scenes and forward-facing capture of bounded and unbounded scenes. NeRF fits multi-layer perceptrons (MLPs) representing view-invariant opacity and view-dependent color volumes to a set of training images, and samples novel views based on volume rendering techniques. In this technical report, we first remark on radiance fields and their potential ambiguities, namely the shape-radiance ambiguity, and analyze NeRF's success in avoiding such ambiguities. Second, we address a parametrization issue involved in applying NeRF to 360 captures of objects within large-scale, unbounded 3D scenes. Our method improves view synthesis fidelity in this challenging scenario. Code is available at https://github.com/Kai-46/nerfplusplus.