NoiseSDF2NoiseSDF: Learning Clean Neural Fields from Noisy Supervision

TL;DR

This paper introduces NoiseSDF2NoiseSDF, a novel method that learns clean neural surface representations directly from noisy point clouds by extending the Noise2Noise paradigm to 3D, improving reconstruction quality.

Contribution

It extends Noise2Noise to 3D neural fields, enabling direct learning of clean neural SDFs from noisy point clouds without clean supervision.

Findings

Significantly improves surface reconstruction from noisy data

Effective on multiple benchmark datasets

Reduces noise impact in neural implicit surface learning

Abstract

Reconstructing accurate implicit surface representations from point clouds remains a challenging task, particularly when data is captured using low-quality scanning devices. These point clouds often contain substantial noise, leading to inaccurate surface reconstructions. Inspired by the Noise2Noise paradigm for 2D images, we introduce NoiseSDF2NoiseSDF, a novel method designed to extend this concept to 3D neural fields. Our approach enables learning clean neural SDFs directly from noisy point clouds through noisy supervision by minimizing the MSE loss between noisy SDF representations, allowing the network to implicitly denoise and refine surface estimations. We evaluate the effectiveness of NoiseSDF2NoiseSDF on benchmarks, including the ShapeNet, ABC, Famous, and Real datasets. Experimental results demonstrate that our framework significantly improves surface reconstruction quality from noisy inputs.