soccernetpro-localization-tennis
收藏资源简介:
SoccerNetPro Localization (Tennis) 是一个专注于网球动作识别与时间定位的数据集,采用 OpenSportsLab / SoccerNet 风格的数据格式。该数据集包含数千个短视频片段(.mp4 格式),并按训练集、验证集和测试集进行组织。每个视频片段配有相应的 JSON 格式的定位标注,标注内容包括事件类型、时间戳(以毫秒为单位,相对于片段的起始时间)以及可选的元数据(如帧率、视频宽高等)。数据集适用于视频分类、时间动作定位等任务,特别适合体育视频分析领域的研究。数据来源于 ECCV 2022 论文《Spotting Temporally Precise, Fine-Grained Events in Video》中发布的原始网球数据。
SoccerNetPro Localization (Tennis) is a dataset dedicated to tennis action recognition and temporal localization, adopting the data format specification of OpenSportsLab / SoccerNet. This dataset comprises thousands of short video clips in .mp4 format, which are partitioned into training, validation, and test splits. Each clip is paired with corresponding localization annotations in JSON format, containing event type, timestamp (in milliseconds relative to the clip's start time), and optional metadata such as frame rate, video resolution (width and height), etc. This dataset is applicable to tasks including video classification and temporal action localization, and is particularly well-suited for research in the domain of sports video analysis. The dataset originates from the original tennis data released in the ECCV 2022 paper titled "Spotting Temporally Precise, Fine-Grained Events in Video".



