UBI-Fights 异常检测数据集

UBI-Fights 数据集可用于特定的异常检测，并且涵盖广泛和多样的打架场景。该数据集是一个大规模数据集，包括 80 小时的视频，在帧一级完全标注。该数据集由 1,000 个视频组成，其中 216 个视频包含一个打架事件，784 个是正常的日常生活情况。所有可能干扰学习过程的不必要的视频片段（如视频介绍、新闻等）都已被删除。

The UBI-Fights dataset is designed for targeted anomaly detection, covering a broad and diverse range of fighting scenarios. It is a large-scale dataset containing 80 hours of videos that are fully annotated at the frame level. The dataset consists of 1,000 video clips, among which 216 contain fighting incidents and the remaining 784 depict normal daily life situations. All unnecessary video segments that may interfere with the learning process (such as video introductions, news segments, etc.) have been removed.

搜集汇总

背景与挑战

背景概述

UBI-Fights是一个大规模异常检测数据集，包含80小时的视频，涵盖216个打架事件视频和784个正常生活场景视频，所有视频在帧级别进行了完整标注。该数据集专为打架场景的异常检测设计，已去除不必要的干扰片段。