MUlti-Shot EventS (MUSES)

MUSES数据集是由华中科技大学、阿里巴巴集团和牛津大学合作创建的大型视频数据集，专注于多镜头时间事件定位。该数据集包含31,477个事件实例，覆盖716小时视频，每个实例平均包含19个镜头，每部视频平均包含176个镜头。MUSES数据集的特点是频繁的镜头切换，导致事件内部变化大，为当前的时间动作定位方法带来挑战。该数据集适用于训练深度学习模型，旨在解决视频内容自动生成中的材料提取效率问题，特别适用于电视剧和电影中的事件定位。

The MUSES Dataset is a large-scale video dataset co-created by Huazhong University of Science and Technology, Alibaba Group, and the University of Oxford, focusing on multi-shot temporal event localization. It contains 31,477 event instances, covering 716 hours of video. Each event instance contains an average of 19 shots, and each video contains an average of 176 shots. The MUSES Dataset is characterized by frequent shot transitions, which lead to large internal changes of events, posing challenges to current temporal action localization methods. This dataset is suitable for training deep learning models, aiming to solve the problem of material extraction efficiency in automatic video content generation, and is particularly applicable to event localization in TV dramas and movies.