Raqabah: A Multi-Severity Car Accident Image Classification Dataset

This dataset was curated to support the development of AI-powered incident detection systems (such as the Raqabah platform). It is specifically structured for Image Classification tasks, aiming to categorize road scenes into different accident severity levels under various environmental conditions. The dataset combines both synthetic data and real-world CCTV footage to maximize model robustness. Dataset Structure & Classes:1. Normal: Standard traffic flow and safe driving environments.2. Minor: Low-impact incidents, minor bumps, or scratches.3. Moderate: Visible vehicle damage or multi-car scrapes.4. Severe: High-impact crashes, including sub-scenarios for Flipped (overturned) and Burned (on fire) vehicles. Environmental Conditions:Images are captured or generated across multiple atmospheric and lighting variations (Day, Night, Fog, Rain, Dust, and Random backgrounds).本数据集专为支撑AI驱动的事件检测系统（如Raqabah平台）的研发而构建，其专门针对图像分类（Image Classification）任务设计，目标是将不同环境条件下的道路场景按事故严重等级进行分类。本数据集融合了合成数据与真实闭路电视（CCTV）监控录像，以最大化模型的鲁棒性。 数据集结构与类别： 1. 正常（Normal）：标准交通流与安全驾驶环境。 2. 轻微事故（Minor）：低影响事件、轻微剐蹭或划痕。 3. 中度事故（Moderate）：可见车辆受损或多车剐蹭。 4. 严重事故（Severe）：高烈度碰撞场景，包含车辆倾覆（Flipped，overturned）与车辆着火（Burned，on fire）两类子场景。 环境条件： 图像的采集或生成覆盖了多种大气与光照变化场景，涵盖白天、夜间、雾天、雨天、扬尘以及随机背景。