Wang Chun-Yi, Lai Kalin Guanlun, Huang Hsu-Chun, Lin Wei-Ting
Office of Physical Education, National Taichung University of Science and Technology, Taichung, Taiwan.
Department of Computer Science & Information Engineering, National Kaohsiung University of Science and Technology, Kaohsiung, Taiwan.
Data Brief. 2024 Jun 22;55:110665. doi: 10.1016/j.dib.2024.110665. eCollection 2024 Aug.
Tennis is a popular sport, and integrating modern technological advancements can greatly enhance player training. Human pose estimation has seen substantial developments recently, driven by progress in deep learning. The dataset described in this paper was compiled from videos of researchers' friend playing tennis. These videos were retrieved frame by frame to categorize various tennis movements, and human skeleton joints were annotated using COCO-Annotator to generate labelled JSON files. By combining these JSON files with the classified image set, we constructed the dataset for this paper. This dataset enables the training and validation of four tennis postures, forehand shot, backhand shot, ready position, and serves, using deep learning models (such as OpenPose). The researchers believe that this dataset will be a valuable asset to the tennis community and human pose estimation field, fostering innovation and excellence in the sport.
网球是一项广受欢迎的运动，整合现代技术进步能够极大地提升球员训练水平。受深度学习进展的推动，人体姿态估计最近取得了重大进展。本文所述的数据集是从研究人员朋友打网球的视频中汇编而来的。这些视频逐帧检索，以对各种网球动作进行分类，并使用COCO注释器对人体骨骼关节进行注释，以生成带标签的JSON文件。通过将这些JSON文件与分类图像集相结合，我们构建了本文的数据集。该数据集能够使用深度学习模型（如OpenPose）对四种网球姿势进行训练和验证，即正手击球、反手击球、准备姿势和发球。研究人员认为，该数据集将成为网球界和人体姿态估计领域的宝贵资产，促进这项运动的创新和卓越发展。
