FineDiving-Pose Dataset
收藏资源简介:
FineDiving-Pose Dataset是由北京邮电大学网络与交换技术国家重点实验室创建的一个用于动作质量评估(AQA)的数据集,特别针对跳水运动中的姿态估计。该数据集包含12,722条人工标注的姿态标签和288,000条通过自动标注管道生成的姿态标签,总计300,722条数据。数据集的创建过程结合了人工标注和自动生成技术,旨在提高现有低质量人体姿态标签的精度。数据集的应用领域主要集中在计算机视觉中的动作质量评估,特别是跳水等体育运动的自动化评分和姿态分析,旨在解决现有方法在捕捉细微姿态差异和动作连续性方面的不足。
The FineDiving-Pose Dataset is a dataset for Action Quality Assessment (AQA) with a particular focus on pose estimation in diving sports, developed by the State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications. It encompasses a total of 300,722 pose labels, including 12,722 manually annotated ones and 288,000 labels generated through an automatic annotation pipeline. The dataset was constructed by combining manual annotation and automatic generation technologies, with the goal of enhancing the precision of existing low-quality human pose labels. Its primary application domains lie in action quality assessment within computer vision, specifically automated scoring and pose analysis for sports such as diving, aiming to mitigate the limitations of existing methods in capturing subtle pose differences and motion continuity.




