NoiseSDF2NoiseSDF: Learning Clean Neural Fields from Noisy Supervision
This addresses the challenge of accurate 3D surface reconstruction from low-quality scans, which is important for applications in computer vision and graphics, but it is incremental as it extends the Noise2Noise paradigm to 3D neural fields.
The paper tackles the problem of reconstructing accurate implicit surfaces from noisy point clouds by introducing NoiseSDF2NoiseSDF, a method that learns clean neural SDFs directly from noisy supervision, and it demonstrates significant improvements in surface reconstruction quality on benchmarks like ShapeNet, ABC, Famous, and Real datasets.
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.