CUHK Avenue

该数据集设置在校园环境中，涵盖单一场景，记录了诸如突然奔跑、抛掷背包和嬉戏行为等异常情况。值得注意的是，数据集采用了正面摄像头角度，人群相对密集，且存在较大程度的遮挡。规模上，这是一个单一场景的数据集，其任务专注于视频异常检测。

This dataset is constructed in a campus environment with a single fixed scene, and documents abnormal behaviors such as sudden running, backpack throwing, and playful frolicking. Notably, the dataset employs a front-facing camera setup, features a relatively dense crowd, and exhibits a high degree of occlusions. As a single-scene dataset, it is specifically tailored for the task of video anomaly detection.

搜集汇总

背景与挑战

背景概述

CUHK Avenue数据集是一个用于异常事件检测的视频数据集，包含16个训练视频和21个测试视频，总计30652帧。训练视频仅包含正常情况，而测试视频则包含正常和异常事件，适用于异常事件检测算法的开发和评估。