Traffic Accident Detection

A new dataset of diverse traffic accidents. Source: Unsupervised Traffic Accident Detection in First-Person Videos

Car Crash Dataset (CCD) is collected for traffic accident analysis. It contains real traffic accident videos captured by dashcam mounted on driving vehicles, which is critical to developing safety-guaranteed self-driving systems. CCD is distinguished from existing datasets for diversified accident annotations, including environmental attributes (day/night, snowy/rainy/good weather conditions), whether ego-vehicles involved, accident participants, and accident reason descriptions. Source: https://github.com/Cogito2012/CarCrashDataset Image Source: https://github.com/Cogito2012/CarCrashDataset

The Traffic Accident Prediction (TAP) data repository offers extensive coverage for 1,000 US cities (TAP-city) and 49 states (TAP-state), providing real-world road structure data that can be easily used for graph-based machine learning methods such as Graph Neural Networks. Additionally, it features multi-dimensional geospatial attributes, including angular and directional features, that are useful for analyzing transportation networks. The TAP repository has the potential to benefit the research community in various applications, including traffic crash prediction, road safety analysis, and traffic crash mitigation. The datasets can be accessed in the TAP-city and TAP-state directories. For example, this repository can aid in traffic accident occurrence prediction and accident severity prediction. Binary labels are used to indicate whether a node contains at least one accident for the occurrence prediction task, while severity is represented by a number between 0 and 7 for the severity prediction task. A severity level of 0 denotes no accident, and 1 to 7 represents increasingly significant impacts on traffic.

CAP-DATA is a large-scale benchmark consisting of 11,727 in-the-wild accident videos with over 2.19 million frames together with labeled fact-effect-reason-introspection description and temporal accident frame label. It can support many useful tasks for accident inference, such as accident detection and prediction (AccidentDet/Pre), causal inference of accident (Accident-Causal), accident classification (Accident-Cla), text-video based accident retrieval (Accident-Retri), and question answering in an accident (Accident-QA) of the driving scene. Source: Cognitive Accident Prediction in Driving Scenes: A Multimodality Benchmark Image Source: https://arxiv.org/pdf/2212.09381v1.pdf