闭路电视攻击姿势与打斗检测数据集

1. 简介 闭路电视攻击姿势与打斗检测数据集是由 Simuletic 制作的精选合成图像数据集（约 100 帧），聚焦高角度闭路电视视角的攻击行为与战斗姿势；数据预先标注为 YOLOv8 姿态估计格式（含边界框 + 17 个 COCO 格式骨架关键点），为完全合成数据，可消除真实监控录像的 GDPR / 隐私担忧，旨在支撑公共场所暴力、异常行为的识别模型训练，完整企业数据集可联系 Simuletic 获取

1. Introduction
The CCTV Attack Pose and Fighting Detection Dataset is a curated synthetic image dataset (approximately 100 frames) created by Simuletic, focusing on attack behaviors and combat postures captured from a high-angle CCTV perspective. All data is pre-annotated in the YOLOv8 pose estimation format, including bounding boxes and 17 COCO-format skeletal keypoints. As a fully synthetic dataset, it eliminates GDPR and privacy concerns associated with real surveillance footage, and is designed to support the training of recognition models for detecting violence and abnormal behaviors in public spaces. The full enterprise-scale dataset can be obtained by contacting Simuletic.

搜集汇总

背景与挑战

背景概述

该数据集是一个由Simuletic制作的合成图像数据集，包含约100帧高角度闭路电视视角的图像，专注于攻击姿势与打斗行为的检测。数据预先标注为YOLOv8姿态估计格式，包括边界框和17个COCO格式骨架关键点，旨在支持公共场所暴力识别模型的训练，同时通过合成数据避免真实监控录像的隐私问题。