用噪声点云训练干净的3D神经场，无需真实标签。
NoiseSDF2NoiseSDF: Learning Clean Neural Fields from Noisy Supervision
- 用噪声点云间直接对比学习，隐式去除表面噪声。
- 在ShapeNet、ABC等数据集上，重建精度显著提升。
- 适合低质量扫描设备获取的噪声点云处理。
从点云重建精确的隐式表面表示仍具挑战性，尤其当数据由低质量扫描设备捕获时。这些点云常含大量噪声，导致表面重建不准确。受2D图像去噪范式Noise2Noise启发，我们提出NoiseSDF2NoiseSDF，将该思想扩展至3D神经场。通过最小化噪声SDF表示间的MSE损失，实现从噪声监督中学习干净的神经SDF，使网络隐式完成去噪并优化表面估计。我们在ShapeNet、ABC、Famous和Real等多个基准数据集上验证了该方法的有效性。实验结果表明，该框架能显著提升噪声输入下的表面重建质量。
原文摘要 · Abstract (English)
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 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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。