Weakly Supervised 3D Object Detection via Multi-Level Visual Guidance

TL;DR

This paper introduces a novel weakly supervised 3D object detection framework that leverages multi-level visual constraints from 2D data without requiring any 3D annotations, achieving competitive results on KITTI.

Contribution

The paper proposes a new framework utilizing feature, output, and training constraints to learn 3D detection from 2D labels only, eliminating the need for 3D annotations.

Findings

Achieves competitive performance without 3D labels.

Outperforms some state-of-the-art weakly supervised methods.

Close to fully supervised methods with 500 3D annotations.

Abstract

Weakly supervised 3D object detection aims to learn a 3D detector with lower annotation cost, e.g., 2D labels. Unlike prior work which still relies on few accurate 3D annotations, we propose a framework to study how to leverage constraints between 2D and 3D domains without requiring any 3D labels. Specifically, we employ visual data from three perspectives to establish connections between 2D and 3D domains. First, we design a feature-level constraint to align LiDAR and image features based on object-aware regions. Second, the output-level constraint is developed to enforce the overlap between 2D and projected 3D box estimations. Finally, the training-level constraint is utilized by producing accurate and consistent 3D pseudo-labels that align with the visual data. We conduct extensive experiments on the KITTI dataset to validate the effectiveness of the proposed three constraints. Without using any 3D labels, our method achieves favorable performance against state-of-the-art approaches and is competitive with the method that uses 500-frame 3D annotations. Code will be made publicly available at https://github.com/kuanchihhuang/VG-W3D.