Simuletic/CCTV_Incident_Dataset_Fall_Lying_Down_Detection

这是一个用于计算机视觉任务的开源合成数据集，专门设计用于从CCTV俯视角度进行跌倒检测、姿态估计和事件监控。数据集包含关键点（姿态）注释，使模型能够理解人体姿势并准确区分站立和跌倒的个体。数据集特点包括双注释（边界框和17关键点骨架）、针对性强（约95%的样本为跌倒状态）、隐私优先（完全合成，无真实个体）和兼容性（适用于YOLOv8-Pose和YOLO11-Pose模型）。数据集结构遵循YOLO Pose格式，包含图像和标签文件，类别分为跌倒和站立。

This is an open-source synthetic dataset for Computer Vision (CV) tasks, specifically designed for Fall Detection, Pose Estimation, and Incident Monitoring from overhead CCTV perspectives. The dataset includes Keypoints (Pose) annotations, enabling models to understand human posture and accurately distinguish between standing and fallen individuals. Key features include Dual Annotations (Bounding Boxes and 17-Keypoint Skeletons), Targeted Scenarios (approx. 95% of subjects are laying/fallen), Privacy-First (fully synthetic, no real individuals), and Compatibility (ready for YOLOv8-Pose and YOLO11-Pose models). The dataset follows the YOLO Pose format, containing image and label files, with classes divided into laying and standing.