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

2025-12-05

AI Summary

In unsupervised 3D object detection, severe noise in initial pseudo-labels critically hinders network convergence. To address this, we propose a three-stage collaborative framework: Occupancy-Guided Warm-up, Instance-Cued Reasoning, and Weight-Adapted Self-Training. First, occupancy prediction guides backbone pre-warming to mitigate early pseudo-label bias. Second, leveraging large-model semantic priors, we design an instance-level prompt reasoning module for precise pseudo-label filtering and confidence calibration. Third, a spatially aware dynamic self-training mechanism refines iterative optimization. Evaluated on Waymo Open Dataset and KITTI, our method achieves mAP improvements of over 15.0% against state-of-the-art unsupervised approaches, significantly enhancing both robustness and convergence stability in unsupervised 3D detection.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: 3D Computer Vision
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• Unsupervised 3D object detection reduces annotation costs in autonomous driving
• Existing methods suffer from incorrect pseudo-labels misleading the training process
• Effectively filtering and refining pseudo-labels remains a critical challenge

Innovation

• Occupancy Guided Warm-up initializes backbone with spatial perception
• Instance-Cued Reasoning uses large model priors to filter pseudo-labels
• Weight-adapted Self-training dynamically re-weights pseudo-labels for improvement