LiFlow: Flow Matching for 3D LiDAR Scene Completion

TL;DR

LiFlow introduces a flow matching framework for 3D LiDAR scene completion, addressing limitations of diffusion models by aligning training and inference distributions, leading to improved local and global scene reconstruction.

Contribution

This work is the first to apply flow matching to 3D LiDAR scene completion, enhancing consistency and performance over previous diffusion-based approaches.

Findings

Achieves state-of-the-art results on multiple benchmarks.

Uses a nearest neighbor flow matching loss and Chamfer distance for better alignment.

Outperforms existing methods in local structure and global coverage.

Abstract

In autonomous driving scenarios, the collected LiDAR point clouds can be challenged by occlusion and long-range sparsity, limiting the perception of autonomous driving systems. Scene completion methods can infer the missing parts of incomplete 3D LiDAR scenes. Recent methods adopt local point-level denoising diffusion probabilistic models, which require predicting Gaussian noise, leading to a mismatch between training and inference initial distributions. This paper introduces the first flow matching framework for 3D LiDAR scene completion, improving upon diffusion-based methods by ensuring consistent initial distributions between training and inference. The model employs a nearest neighbor flow matching loss and a Chamfer distance loss to enhance both local structure and global coverage in the alignment of point clouds. LiFlow achieves state-of-the-art performance across multiple metrics. Code: https://github.com/matteandre/LiFlow.