Toward Robust Neural Reconstruction from Sparse Point Sets

TL;DR

This paper introduces a distributionally robust optimization framework for learning signed distance functions from sparse, noisy 3D point clouds, improving stability and accuracy without relying on smoothness priors.

Contribution

It presents a novel DRO-based method with a regularization term that utilizes uncertainty samples, enabling stable, supervised-free SDF learning from sparse data.

Findings

Outperforms existing methods on synthetic and real datasets

Enhances stability and robustness in SDF learning

Effective across multiple data modalities

Abstract

We consider the challenging problem of learning Signed Distance Functions (SDF) from sparse and noisy 3D point clouds. In contrast to recent methods that depend on smoothness priors, our method, rooted in a distributionally robust optimization (DRO) framework, incorporates a regularization term that leverages samples from the uncertainty regions of the model to improve the learned SDFs. Thanks to tractable dual formulations, we show that this framework enables a stable and efficient optimization of SDFs in the absence of ground truth supervision. Using a variety of synthetic and real data evaluations from different modalities, we show that our DRO based learning framework can improve SDF learning with respect to baselines and the state-of-the-art methods.