LOGO

LOGO数据集是由深圳国际研究生院，清华大学创建的多人长格式视频数据集，专注于群体动作质量评估。该数据集包含200个来自26个艺术游泳事件的视频，每个样本平均时长204.2秒，总时长超过11小时。数据集提供了详细的动作和形成标签，用于描述多个运动员之间的群体信息和动作过程。LOGO数据集通过其场景复杂性和丰富的注释，为构建能够模拟演员间群体信息的动作质量评估方法提供了潜力。此外，数据集的长视频时长也挑战了方法聚合长期时间信息的能力。

The LOGO Dataset is a multi-person long-form video dataset created by the Shenzhen International Graduate School, Tsinghua University, focusing on group movement quality assessment. This dataset contains 200 videos from 26 artistic swimming events, with an average duration of 204.2 seconds per sample and a total duration exceeding 11 hours. The dataset provides detailed movement and formation labels to describe group information and movement processes among multiple athletes. Benefiting from its complex scenarios and rich annotations, the LOGO Dataset offers potential for developing movement quality assessment methods that can model group information among athletes. Furthermore, the long-duration videos in the dataset also pose challenges to the ability of methods to aggregate long-term temporal information.