Comprehensive Dataset for AI-Driven Traffic Accident Detection and Computer Vision Systems

We introduce a novel dataset consisting of approximately 5,700 video files, specifically designed to enhance the development of real-time traffic accident detection systems in smart city environments. It encompasses a diverse range of traffic scenarios, captured through Traffic/Surveillance Cameras (Trafficam) and Dash Cameras (Dashcam), along with additional external data sources. The dataset is meticulously organized into three segments: Training, Validation, and Testing, with each segment offering a unique blend of traffic and dashcam footage across different scenarios.The dataset is divided into eight classes: Backend, Backend Rollover, Frontend, Frontend Rollover, No Accident Normal Traffic, Sidehit, Sidehit Rollover, and General Augmented Crash. These classes provide a rich tapestry of real-world situations, ranging from routine traffic conditions to complex accident scenes. The distribution of the dataset is as follows: 3,912 files for Training, 1,054 for Validation, and 725 for Testing, encompassing a mix of accident and normal traffic scenarios from both Trafficam and Dashcam sources, along with additional external data.The videos have been processed and segmented into five-second non-overlapping clips to ensure conciseness, focusing on the rapid dynamics of accidents. This careful curation and classification make the dataset an invaluable resource for training and evaluating AI models in traffic safety applications. By providing a wide array of scenarios, this dataset enables researchers and developers to develop state-of-the-art algorithms, ensuring high accuracy and reliability in diverse urban settings. This dataset is crucial for academic research and also serves as a practical tool for improving traffic management and safety in smart cities, contributing significantly to the collaborative efforts in creating safer, more efficient urban environments.

本研究提出一款新型数据集，包含约5700个视频文件，专为优化智慧城市场景下的实时交通事故检测系统开发打造。该数据集覆盖多样化交通场景，数据采集自交通/监控摄像头（Traffic/Surveillance Cameras，Trafficam）与行车记录仪（Dash Cameras，Dashcam），同时纳入额外外部数据源。数据集经过精细统筹划分，分为训练集（Training）、验证集（Validation）与测试集（Testing）三个子集，每个子集均包含不同场景下的交通摄像头与行车记录仪影像混合样本。本数据集共划分为8个类别：Backend、Backend Rollover、Frontend、Frontend Rollover、No Accident Normal Traffic、Sidehit、Sidehit Rollover 以及 General Augmented Crash。这些类别涵盖了从日常交通工况到复杂事故现场的各类真实场景，内容详实丰富。数据集的样本分布如下：训练集3912个文件，验证集1054个文件，测试集725个文件，涵盖来自Trafficam与Dashcam采集的事故与正常交通场景样本，同时包含额外外部数据源样本。所有视频均经过预处理，被分割为时长5秒的非重叠片段，既保证内容紧凑简洁，又可聚焦事故发生时的快速动态过程。这种精细化的整理与分类工作，使得该数据集成为交通安全应用场景中训练与评估人工智能（AI）模型的宝贵资源。依托丰富多样的场景覆盖，该数据集可支持研究人员与开发者研发前沿算法，保障其在多样化城市场景中具备高精度与高可靠性。本数据集不仅对学术研究具有重要价值，同时可作为提升智慧城市交通管理与安全水平的实用工具，为构建更安全、更高效的城市环境的协同工作提供关键支撑。

搜集汇总

背景与挑战

背景概述

该数据集是一个包含5,700个视频文件的交通视频数据集，专门用于训练和评估AI模型在智能城市环境中的交通事故检测能力。数据集涵盖了多种交通场景，包括正常交通和不同类型的事故，分为训练、验证和测试三个部分，并细分为8个具体类别。