OWL: Unsupervised 3D Object Detection by Occupancy Guided Warm-up and Large Model Priors Reasoning

TL;DR

This paper introduces OWL, an unsupervised 3D object detection framework that uses occupancy-guided warm-up, large-model priors, and dynamic self-training to improve detection accuracy without annotations.

Contribution

OWL combines occupancy-guided warm-up, large-model prior reasoning, and adaptive self-training to enhance unsupervised 3D object detection performance.

Findings

Outperforms state-of-the-art methods by over 15% mAP on WOD and KITTI datasets.

Effectively filters and refines pseudo-labels using large-model priors.

Improves backbone initialization with occupancy-guided warm-up.

Abstract

Unsupervised 3D object detection leverages heuristic algorithms to discover potential objects, offering a promising route to reduce annotation costs in autonomous driving. Existing approaches mainly generate pseudo labels and refine them through self-training iterations. However, these pseudo-labels are often incorrect at the beginning of training, resulting in misleading the optimization process. Moreover, effectively filtering and refining them remains a critical challenge. In this paper, we propose OWL for unsupervised 3D object detection by occupancy guided warm-up and large-model priors reasoning. OWL first employs an Occupancy Guided Warm-up (OGW) strategy to initialize the backbone weight with spatial perception capabilities, mitigating the interference of incorrect pseudo-labels on network convergence. Furthermore, OWL introduces an Instance-Cued Reasoning (ICR) module that leverages the prior knowledge of large models to assess pseudo-label quality, enabling precise filtering and refinement. Finally, we design a Weight-adapted Self-training (WAS) strategy to dynamically re-weight pseudo-labels, improving the performance through self-training. Extensive experiments on Waymo Open Dataset (WOD) and KITTI demonstrate that OWL outperforms state-of-the-art unsupervised methods by over 15.0% mAP, revealing the effectiveness of our method.