NoiseSDF2NoiseSDF: Learning Clean Neural Fields from Noisy Supervision

Tengkai Wang, Weihao Li, Ruikai Cui, Shi Qiu, Nick Barnes

Keywords: Noise2Noise;Neural Field;SDF;Point Cloud;Surface Reconstruction

TL;DR: Use 2D Noise2Noise Framework to denoise neural fields and denoising neural SDF to recover clean surface as an example

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 RealWorld datasets. Experimental results demonstrate that our framework significantly improves surface reconstruction quality from noisy inputs.

Primary Area: Deep learning (e.g., architectures, generative models, optimization for deep networks, foundation models, LLMs)

Submission Number: 12064