S3PT: Scene Semantics and Structure Guided Clustering to Boost Self-Supervised Pre-Training for Autonomous Driving

TL;DR

This paper introduces S3PT, a scene-aware self-supervised pre-training method for autonomous driving that leverages semantic, spatial, and depth information to improve downstream detection and segmentation performance.

Contribution

S3PT proposes scene semantics and structure guided clustering techniques, including semantic distribution, spatial diversity, and depth-guided clustering, to enhance self-supervised pre-training for autonomous driving.

Findings

Improves semantic segmentation accuracy on nuScenes, nuImages, Cityscapes.

Enhances 3D object detection performance.

Shows promising domain translation capabilities.

Abstract

Recent self-supervised clustering-based pre-training techniques like DINO and Cribo have shown impressive results for downstream detection and segmentation tasks. However, real-world applications such as autonomous driving face challenges with imbalanced object class and size distributions and complex scene geometries. In this paper, we propose S3PT a novel scene semantics and structure guided clustering to provide more scene-consistent objectives for self-supervised training. Specifically, our contributions are threefold: First, we incorporate semantic distribution consistent clustering to encourage better representation of rare classes such as motorcycles or animals. Second, we introduce object diversity consistent spatial clustering, to handle imbalanced and diverse object sizes, ranging from large background areas to small objects such as pedestrians and traffic signs. Third, we propose a depth-guided spatial clustering to regularize learning based on geometric information of the scene, thus further refining region separation on the feature level. Our learned representations significantly improve performance in downstream semantic segmentation and 3D object detection tasks on the nuScenes, nuImages, and Cityscapes datasets and show promising domain translation properties.