Malta Audio-Visual Anomaly Detection (MAVAD)
收藏资源简介:
Malta Audio-Visual Anomaly Detection (MAVAD)数据集是由奥胡斯大学数字技术部创建的,用于交通视频中的异常检测。该数据集包含764个视频,涵盖多种天气和光照条件,从马耳他岛的三个不同地点收集。数据集通过模糊人脸和车牌来保护个人隐私,并包含11个类别,如行人、公交车和重型车辆等。创建过程涉及部署音频-视觉摄像头,并使用手工制作的掩码来防止记录私人财产。该数据集主要应用于视频异常检测,旨在通过结合视觉和音频特征提高异常检测的准确性。
The Malta Audio-Visual Anomaly Detection (MAVAD) Dataset was developed by the Department of Digital Technology, Aarhus University, for anomaly detection in traffic videos. Comprising 764 videos collected from three distinct locations on the island of Malta, this dataset covers a wide range of weather and lighting conditions. Personal privacy is protected via blurring of faces and license plates, and the dataset includes 11 categories including pedestrians, buses, and heavy-duty vehicles. During the dataset creation process, audio-visual cameras were deployed, and handcrafted masks were utilized to avoid recording private property. This dataset is primarily applied to video anomaly detection, aiming to improve the accuracy of anomaly detection by combining visual and audio features.




