Rethinking the Encoding and Annotating of 3D Bounding Box: Corner-Aware 3D Object Detection from Point Clouds

2025-11-18

AI Summary

Center-aligned regression in LiDAR point cloud 3D detection suffers from unstable localization and high prediction noise, as object centers often fall in sparse or empty regions of the Bird’s Eye View (BEV) feature map. Method: This paper proposes a novel corner-aligned regression paradigm, replacing the geometrically ambiguous center with geometrically identifiable 3D bounding box corners as regression targets—leveraging dense, reliably observable BEV regions to enhance robustness. It introduces corner supervision into weakly supervised 3D detection for the first time, integrating geometric constraints and 2D image bounding box priors; training requires only BEV corner annotations, drastically reducing annotation cost. A corner-aware detection head is designed to support partial parameter recovery and seamless integration with mainstream BEV-based detectors. Results: On KITTI, the method achieves a 3.5% AP improvement over center-aligned baselines and attains 83% of the performance of fully supervised methods using corner-only annotations.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Hardware-aware ML
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

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

Abstract

Center-aligned regression remains dominant in LiDAR-based 3D object detection, yet it suffers from fundamental instability: object centers often fall in sparse or empty regions of the bird's-eye-view (BEV) due to the front-surface-biased nature of LiDAR point clouds, leading to noisy and inaccurate bounding box predictions. To circumvent this limitation, we revisit bounding box representation and propose corner-aligned regression, which shifts the prediction target from unstable centers to geometrically informative corners that reside in dense, observable regions. Leveraging the inherent geometric constraints among corners and image 2D boxes, partial parameters of 3D bounding boxes can be recovered from corner annotations, enabling a weakly supervised paradigm without requiring complete 3D labels. We design a simple yet effective corner-aware detection head that can be plugged into existing detectors. Experiments on KITTI show our method improves performance by 3.5% AP over center-based baseline, and achieves 83% of fully supervised accuracy using only BEV corner clicks, demonstrating the effectiveness of our corner-aware regression strategy.

Problem

• Addresses instability in center-based 3D object detection from LiDAR point clouds
• Proposes corner-aligned regression using geometrically stable corner features
• Enables weakly supervised detection with partial annotations via corner constraints

Innovation

• Corner-aligned regression replaces center-based prediction
• Geometric constraints enable weakly supervised learning
• Plug-in detection head enhances existing 3D detectors