4D Contrastive Superflows are Dense 3D Representation Learners

TL;DR

SuperFlow is a novel 3D representation learning framework for autonomous driving that leverages spatiotemporal data from LiDAR-camera pairs, using dense-to-sparse regularization and flow-based contrastive learning to improve perception models.

Contribution

It introduces SuperFlow, a new framework combining dense-to-sparse regularization and flow-based contrastive learning for effective 3D perception pretraining.

Findings

Outperforms existing methods across 11 datasets

Enhances 3D perception accuracy and robustness

Reveals properties of scaling backbone models

Abstract

In the realm of autonomous driving, accurate 3D perception is the foundation. However, developing such models relies on extensive human annotations -- a process that is both costly and labor-intensive. To address this challenge from a data representation learning perspective, we introduce SuperFlow, a novel framework designed to harness consecutive LiDAR-camera pairs for establishing spatiotemporal pretraining objectives. SuperFlow stands out by integrating two key designs: 1) a dense-to-sparse consistency regularization, which promotes insensitivity to point cloud density variations during feature learning, and 2) a flow-based contrastive learning module, carefully crafted to extract meaningful temporal cues from readily available sensor calibrations. To further boost learning efficiency, we incorporate a plug-and-play view consistency module that enhances the alignment of the knowledge distilled from camera views. Extensive comparative and ablation studies across 11 heterogeneous LiDAR datasets validate our effectiveness and superiority. Additionally, we observe several interesting emerging properties by scaling up the 2D and 3D backbones during pretraining, shedding light on the future research of 3D foundation models for LiDAR-based perception.