EC-IoU: Orienting Safety for Object Detectors via Ego-Centric Intersection-over-Union

2024-03-20

IEEE/RJS International Conference on Intelligent RObots and Systems

AI Summary

To address the dual challenges of insufficient detector safety assurance and poor detection performance for nearby objects from the ego-vehicle’s perspective in autonomous driving, this paper proposes EC-IoU (Ego-Centric IoU), a novel evaluation metric and differentiable loss function. EC-IoU introduces a viewpoint-aware spatial weighting scheme that dynamically modulates IoU penalties based on object-to-ego distance, prioritizing precise localization and coverage of proximal, safety-critical targets. This represents the first shift from conventional geometric consistency–based detection evaluation to an ego-centric consistency paradigm explicitly aligned with safety-critical distance requirements. Integrated into an end-to-end optimization framework on the KITTI dataset, EC-IoU yields detectors with higher mean Average Precision (mAP) than standard IoU baselines, significantly improved recall for near-range objects, enhanced reliability of collision warning systems, and demonstrably stronger support for downstream safety-critical decision-making tasks.

Technology Category

• Computer Vision: Vision for Robotics & Autonomous Driving
• Search and Optimization: Evaluation and Analysis
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

This paper presents Ego-Centric Intersection-over-Union (EC-IoU), addressing the limitation of the standard IoU measure in characterizing safety-related performance for object detectors in navigating contexts. Concretely, we propose a weighting mechanism to refine IoU, allowing it to assign a higher score to a prediction that covers closer points of a ground-truth object from the ego agent’s perspective. The proposed EC-IoU measure can be used in typical evaluation processes to select object detectors with better safety-related performance for downstream tasks. It can also be integrated into common loss functions for model fine-tuning. While geared towards safety, our experiment with the KITTI dataset demonstrates the performance of a model trained on EC-IoU can be better than that of a variant trained on IoU in terms of mean Average Precision as well.

Problem

• Object Detection
• First-Person Perspective
• Driving Safety

Innovation

• EC-IoU
• Object Detection
• Self-Centric Perspective