Shanghai University of Sport (SUS), in collaboration with Shanghai University and Chinese tech giant Baidu, has recently launched an AI model for the sports industry, the first of its kind in China.
The model comprises three vertical large models. The first one gleans information from sports literature worldwide to answer sports-related questions.
The second one specializes in motion recognition and tactical analysis. It can automatically analyze videos and images from sports training and produce detailed analyses of human postures in 2D and 3D. It can also provide quantifiable indicators essential for further bio-mechanical assessments, such as distance, speed, height, and angular velocity.
Currently, this model is used by Chinese national teams in different sports including diving, swimming, track and field, gymnastics and rock climbing. It is also being used to assist athletes in their daily training and preparations for the 2024 Paris Olympics.
The third one is a large multimodal model, which integrates text, audio, video, 3D, and analysis by the other two models to answer sports-related questions, analyze technical movements, generate AI commentary and offer personalized courses.
In an interview with a local media, Professor Li Yongming of SUS explained how the AI models can improve the performance of athletes. In sports such as diving or swimming, coaches traditionally rely on post-session video analysis. But now AI provides real-time analytical data as soon as an athlete dives into the water or emerges from it, significantly increasing training efficacy.
Similarly, China's national basketball team has partnered with AI software company SenseTime. The latter can construct 3D court models and use 3D motion capture algorithms to analyze the real-time motion status of all players in the court and the trajectory of the ball.
Additionally, the AI model is used for data statistics and game video processing, supporting training strategies and competitive tactics.