Towards Vision Zero: The Accid3nD Dataset

2025-03-15

AI Summary

Existing public datasets lack multimodal 3D annotations of real-world road accidents, hindering accurate accident modeling and analysis. To address this, we introduce Accid3nD—the first multimodal 3D-annotated dataset specifically designed for real highway accidents—comprising synchronized four-view camera streams and 25-Hz LiDAR data across 111,945 frames. It provides 2D/3D bounding boxes, instance masks, and uniquely assigned 3D trajectory IDs (first publicly released), capturing high-dynamic collision scenarios under diverse weather and illumination conditions. We propose a real-time accident detection framework integrating rule-based heuristics with lightweight learning, built upon precise multi-sensor calibration, OpenLABEL-compliant annotation, and end-to-end multi-task modeling. Evaluated on Accid3nD, our method achieves 82.7% mAP@0.5, significantly outperforming purely learning-based approaches. Accid3nD enables downstream applications including accident reconstruction, causal inference, and V2X-enabled early warning. All code, models, and data are fully open-sourced.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multimodal Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Even though a significant amount of work has been done to increase the safety of transportation networks, accidents still occur regularly. They must be understood as unavoidable and sporadic outcomes of traffic networks. No public dataset contains 3D annotations of real-world accidents recorded from roadside sensors. We present the Accid3nD dataset, a collection of real-world highway accidents in different weather and lighting conditions. It contains vehicle crashes at high-speed driving with 2,634,233 labeled 2D bounding boxes, instance masks, and 3D bounding boxes with track IDs. In total, the dataset contains 111,945 labeled frames recorded from four roadside cameras and LiDARs at 25 Hz. The dataset contains six object classes and is provided in the OpenLABEL format. We propose an accident detection model that combines a rule-based approach with a learning-based one. Experiments and ablation studies on our dataset show the robustness of our proposed method. The dataset, model, and code are available on our website: https://accident-dataset.github.io.

Problem

• Lack of public dataset with 3D annotations for real-world accidents.
• Need for understanding accidents in diverse weather and lighting conditions.
• Development of a robust accident detection model combining rule-based and learning-based approaches.

Innovation

• Accid3nD dataset with 3D annotations
• Combines rule-based and learning-based detection
• OpenLABEL format with six object classes