Unveiling the Ambiguity in Neural Inverse Rendering: A Parameter Compensation Analysis

TL;DR

This paper investigates the inherent ambiguity in neural inverse rendering, demonstrating how scene property estimations can compensate for each other, and emphasizes the need for additional priors to improve accuracy.

Contribution

It introduces an evaluation framework and a disentangled neural microfacet field model to analyze and mitigate ambiguity in neural inverse rendering.

Findings

Intrinsic ambiguity exists in neural inverse rendering.

Adjusting one scene property can compensate for perturbations in another.

Additional priors are crucial for accurate scene reconstruction.

Abstract

Inverse rendering aims to reconstruct the scene properties of objects solely from multiview images. However, it is an ill-posed problem prone to producing ambiguous estimations deviating from physically accurate representations. In this paper, we utilize Neural Microfacet Fields (NMF), a state-of-the-art neural inverse rendering method to illustrate the inherent ambiguity. We propose an evaluation framework to assess the degree of compensation or interaction between the estimated scene properties, aiming to explore the mechanisms behind this ill-posed problem and potential mitigation strategies. Specifically, we introduce artificial perturbations to one scene property and examine how adjusting another property can compensate for these perturbations. To facilitate such experiments, we introduce a disentangled NMF where material properties are independent. The experimental findings underscore the intrinsic ambiguity present in neural inverse rendering and highlight the importance of providing additional guidance through geometry, material, and illumination priors.