RWF-2000

该数据集名为RWF-2000，包含了来自不同监控摄像头来源的2000个视频，其中1000个为暴力视频，另外1000个为非暴力视频。所有视频的时长均为5秒，以30帧每秒的速度拍摄。训练该数据集需要大量的RAM，即便是在像Google Colab这样具有高内存资源的平台上也是如此。该数据集的规模为2000个视频，其任务是进行视频中的暴力行为检测。

This dataset is named RWF-2000. It contains 2000 videos collected from various surveillance camera feeds, with 1000 annotated as violent and the remaining 1000 as non-violent. All videos have a fixed duration of 5 seconds and are recorded at 30 frames per second (fps). Training with this dataset requires substantial RAM, even on high-memory platforms such as Google Colab. With a total of 2000 video samples, the core task of this dataset is violent behavior detection in video content.

搜集汇总

背景与挑战

背景概述

RWF-2000是一个包含2000个视频片段的大规模数据库，专门用于暴力行为检测。这些视频片段来源于监控摄像头，经过预处理和标记，但由于隐私问题，视频文件未公开。