Human-related Anomaly Localization dataset
收藏资源简介:
Human-related Anomaly Localization dataset是由北京邮电大学国家网络与交换技术重点实验室收集和注释的,该数据集包含12种类型的人类相关异常行为,如打斗、跌倒、抢劫等,共1159个视频,超过2543000帧,每个视频都有帧级注释的异常区间以及相应的帧标题和逻辑链文本。该数据集用于扩展少样本动作定位任务的应用范围,解决人类异常行为定位问题。
The Human-related Anomaly Localization dataset was collected and annotated by the State Key Laboratory of Networking and Switching Technology at Beijing University of Posts and Telecommunications. This dataset encompasses 12 types of human-related anomalous behaviors, such as fighting, falling, robbery and others, with a total of 1,159 videos and over 2,543,000 frames. Each video is equipped with frame-level annotations of anomalous intervals, along with corresponding frame titles and logical chain texts. This dataset is designed to expand the application scope of few-shot action localization tasks and address the problem of human anomalous behavior localization.

- 1Chain-of-Thought Textual Reasoning for Few-shot Temporal Action Localization北京邮电大学 国家网络与交换技术重点实验室 · 2025年



