ComplexVAD
收藏资源简介:
ComplexVAD是一个大规模视频异常检测数据集,由南佛罗里达大学和三菱电机研究实验室共同创建,旨在解决现有数据集中复杂异常不足的问题。该数据集包含104个训练视频和113个测试视频,总计3,681,438帧,分辨率为1080x1920,帧率为30帧/秒。数据集记录了大学校园内的复杂场景,包括行人、自行车、汽车等多种对象的交互异常。数据集的创建过程历时5个月,涵盖了不同时间段和场景变化。ComplexVAD主要用于视频监控领域,旨在通过建模对象之间的交互来检测复杂异常,提升异常检测的准确性和实用性。
ComplexVAD is a large-scale video anomaly detection dataset jointly created by the University of South Florida and Mitsubishi Electric Research Laboratories, aiming to address the insufficient complex anomalies in existing datasets. This dataset consists of 104 training videos and 113 test videos, with a total of 3,681,438 frames, a resolution of 1080x1920, and a frame rate of 30 frames per second. The dataset records complex scenarios on university campuses, including interactive anomalies among multiple objects such as pedestrians, bicycles, and cars. The dataset was developed over a period of 5 months, covering different time periods and scene variations. ComplexVAD is primarily used in the field of video surveillance, aiming to detect complex anomalies by modeling the interactions between objects, thereby improving the accuracy and practicality of anomaly detection.




