Self-Supervised Robust Scene Flow Estimation via the Alignment of Probability Density Functions

TL;DR

This paper introduces a self-supervised scene flow estimation method that models point clouds as probability density functions, enabling robust and accurate motion recovery even with missing data and outliers.

Contribution

The novel approach converts scene flow estimation into aligning probability density functions using Gaussian mixture models and Cauchy-Schwarz divergence, improving robustness over existing methods.

Findings

Outperforms Chamfer Distance and Earth Mover's Distance in real-world tests

Achieves state-of-the-art results among self-supervised methods on FlyingThings3D and KITTI

Surpasses some supervised methods without requiring ground truth annotations

Abstract

In this paper, we present a new self-supervised scene flow estimation approach for a pair of consecutive point clouds. The key idea of our approach is to represent discrete point clouds as continuous probability density functions using Gaussian mixture models. Scene flow estimation is therefore converted into the problem of recovering motion from the alignment of probability density functions, which we achieve using a closed-form expression of the classic Cauchy-Schwarz divergence. Unlike existing nearest-neighbor-based approaches that use hard pairwise correspondences, our proposed approach establishes soft and implicit point correspondences between point clouds and generates more robust and accurate scene flow in the presence of missing correspondences and outliers. Comprehensive experiments show that our method makes noticeable gains over the Chamfer Distance and the Earth Mover's Distance in real-world environments and achieves state-of-the-art performance among self-supervised learning methods on FlyingThings3D and KITTI, even outperforming some supervised methods with ground truth annotations.