Charlotte Anomaly Dataset (CHAD)
收藏资源简介:
Charlotte Anomaly Dataset (CHAD)是由北卡罗来纳大学夏洛特分校创建的一个高分辨率、多摄像头异常检测数据集,专门设计用于商业停车场环境。CHAD包含超过1.15百万帧的视频数据,首次在异常检测数据集中引入了边界框、身份和姿态标注,特别适用于基于骨架的异常检测。数据集通过四个摄像头的视角捕捉同一场景,支持智能视频监控应用。CHAD不仅规模庞大,还包含了超过1百万帧的正常行为数据,适用于无监督学习方法。数据集的创建过程考虑了实际环境的真实性,确保训练数据能准确代表目标应用场景。CHAD的应用领域主要集中在视频异常检测,旨在解决停车场等场景中的异常行为识别问题。
Charlotte Anomaly Dataset (CHAD) is a high-resolution, multi-camera anomaly detection dataset created by the University of North Carolina at Charlotte, specifically designed for commercial parking lot environments. CHAD contains over 1.15 million video frames, and for the first time introduces bounding box, identity and pose annotations into anomaly detection datasets, making it particularly suitable for skeleton-based anomaly detection. The dataset captures the same scene from four camera perspectives, supporting intelligent video surveillance applications. Beyond its large scale, CHAD also includes over 1 million frames of normal behavior data, which is applicable to unsupervised learning methods. The dataset was developed with consideration of the realism of real-world environments, ensuring that the training data can accurately represent the target application scenario. The main application field of CHAD is video anomaly detection, aiming to solve the problem of abnormal behavior recognition in scenarios such as parking lots.

- 1CHAD: Charlotte Anomaly Dataset北卡罗来纳大学夏洛特分校 · 2023年



