CornerPoint3D: Look at the Nearest Corner Instead of the Center

2025-04-03

AI Summary

To address inaccurate center localization, poor size generalization, and overfitting of conventional evaluation metrics in cross-domain LiDAR 3D detection, this paper proposes a novel detection paradigm that replaces object centers with **nearest corners**, focusing on the sensor-proximal surface—a more robust and observable geometric primitive for collision-avoidance applications. Our contributions are threefold: (1) the first nearest-corner detection framework, built upon CenterPoint, incorporating corner heatmap supervision and a geometry-aware regression loss; (2) a learnable surface-guidance mechanism that fuses BEV features with an EdgeHead refinement head to enhance proximal-surface modeling; and (3) two novel cross-domain evaluation metrics specifically designed for proximal-surface accuracy. Experiments demonstrate consistent superiority over CenterPoint across multiple cross-domain benchmarks, with significant improvements under both the proposed metrics and standard BEV/3D detection metrics—achieving enhanced localization robustness and practical deployment value.

Technology Category

• Computer Vision: 3D Computer Vision
• Search and Optimization: Learning to Search
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

3D object detection aims to predict object centers, dimensions, and rotations from LiDAR point clouds. Despite its simplicity, LiDAR captures only the near side of objects, making center-based detectors prone to poor localization accuracy in cross-domain tasks with varying point distributions. Meanwhile, existing evaluation metrics designed for single-domain assessment also suffer from overfitting due to dataset-specific size variations. A key question arises: Do we really need models to maintain excellent performance in the entire 3D bounding boxes after being applied across domains? Actually, one of our main focuses is on preventing collisions between vehicles and other obstacles, especially in cross-domain scenarios where correctly predicting the sizes is much more difficult. To address these issues, we rethink cross-domain 3D object detection from a practical perspective. We propose two new metrics that evaluate a model's ability to detect objects' closer-surfaces to the LiDAR sensor. Additionally, we introduce EdgeHead, a refinement head that guides models to focus more on learnable closer surfaces, significantly improving cross-domain performance under both our new and traditional BEV/3D metrics. Furthermore, we argue that predicting the nearest corner rather than the object center enhances robustness. We propose a novel 3D object detector, coined as CornerPoint3D, which is built upon CenterPoint and uses heatmaps to supervise the learning and detection of the nearest corner of each object. Our proposed methods realize a balanced trade-off between the detection quality of entire bounding boxes and the locating accuracy of closer surfaces to the LiDAR sensor, outperforming the traditional center-based detector CenterPoint in multiple cross-domain tasks and providing a more practically reasonable and robust cross-domain 3D object detection solution.

Problem

• Improving cross-domain 3D object detection accuracy
• Reducing reliance on object center prediction
• Enhancing collision avoidance in varying point distributions

Innovation

• EdgeHead focuses on learnable closer surfaces
• CornerPoint3D predicts nearest corner not center
• New metrics evaluate closer-surface detection ability