WF数据集

WF数据集是由伦斯勒理工学院电气、计算机与系统工程系创建，专注于长期视频异常检测。该数据集包含2000个1920×1080分辨率的视频，每个视频约1分钟长，采样率为15帧每秒，覆盖了三个月的时间周期。数据集中的视频来自一个公共网络摄像头，记录了一个棒球体育场的活动，包括不同时间、天气和人群条件下的场景。此外，数据集还收集了相关的元数据，如一天中的时间、星期几和预定比赛时间。该数据集旨在解决视频异常检测中的长期上下文感知问题，特别是在复杂的比赛日程和可能拥挤的场景中。

The WF dataset was developed by the Department of Electrical, Computer, and Systems Engineering at Rensselaer Polytechnic Institute, with a core focus on long-term video anomaly detection. It comprises 2000 videos with a resolution of 1920×1080, each lasting approximately 1 minute and recorded at a frame sampling rate of 15 frames per second, spanning a three-month data collection period. The video footage is sourced from a public web camera, capturing on-site activities at a baseball stadium across varying time windows, weather conditions, and crowd densities. Additionally, relevant metadata is collected, including the time of day, day of the week, and scheduled game times. This dataset is designed to address the long-term context-aware problem in video anomaly detection, particularly in scenarios with complex game schedules and potentially crowded environments.