Zhejiang University: ZJU scientists provide technological support for Chinese national table tennis team to repeat glory at Tokyo Olympics
Chinese national table tennis team dominated the Olympic stage with four gold medals and three silver medals in Tokyo. Behind the dominance with a near-perfect medal sweep were players’ rigorous training, coaches’ wisdom and technological support.
On July 27, the semi-final match between China’s SUN Yingsha and Japan’s Mima Ito was in full swing at the Tokyo Metropolitan Gymnasium. You wouldn’t have guessed that besides players, umpires and the few spectators in the stands, a powerful AI platform was monitoring the match.
Every serve, swing and movement of Mima Ito was captured by this AI cloud platform deployed in Tokyo. These data were transmitted at an average speed of 100 Mbps to the technical team of Zhejiang University Table Tennis Intelligent Big Data Analysis Platform, 2,442 kilometers away from the Tokyo Metropolitan Gymnasium.
After the match, a technical and tactical report was fed to the tablets of the coaches and players by the AI cloud platform.
Prof. WU Yingcai, vice dean of the Zhejiang University College of Computer Science and Technology, is the main developer of this platform. The leader of the Zhejiang University Sports Big Data Innovation team is Prof. ZHANG Hui, the scientific researcher on the national table tennis team in the Tokyo Olympics and head of the Department of Sports Science at the Zhejiang University College of Education. This team has been providing the national table tennis team with strong support in data analysis and tactical research.
Compared with other ball games, table tennis involves considerable complex and delicate techniques and tactics. Therefore, the analysis of table tennis matches is particularly difficult and important. According to Zhang, table tennis data used to be collected from videotapes, which sometimes took about five hours.
It is extremely challenging to remove redundant broadcast footage, identify score changes, detect and frame each stroke, and annotate data efficiently and precisely in an automatic or semi-automatic manner on low-quality and low-frame livestreams. Wu led his team to construct an intelligent big data analysis platform for table tennis two years ago. It took the team over one year to develop the platform for table tennis and achieve semi-automatic data annotation on the strength of an interactive visual interface, AI algorithms, and expert knowledge.
This platform has made a series of technological breakthroughs, shortening the labeling time from 2 hours, 1 hour to eventually right after a match ends.
Besides, this platform is able to assess the performance of every stroke. Previously, coaches and researchers needed to manually evaluate the performance of a player’s stroke on site or by watching match videos frame by frame. What if there were 100 or 1,000 videos? This would be an impossible mission, for assessing strokes is highly subjective and depends on expertise knowledge. It thus became a formidable challenge to enable the computer to learn expertise knowledge and make an objective and accurate assessment of each stroke by combining a series of attributes, such as the technique employed in each stroke, the player’s position, and the placement of the ball. Since 2020, Wu has worked in collaboration with Prof. ZHOU Zhihua from Nanjing University to tackle this challenge. By adopting the theoretical framework, abductive learning, they have made significant progress and obtained original findings this year.
“Using the theoretical framework of abductive learning, we fused over 30 technical and tactical rules concerning table tennis into data-driven machine learning models and developed a framework that can automatically assess the performance of every stroke,” said Wu, “As long as videos are uploaded onto the intelligent big data analysis platform, the calculation can be performed automatically, which will be gradually applied to routine training.”
The platform has stored the data of more than 8,000 matches so far. The research team has started to work on match simulation as of 2018. “Using big data and AI, we can precisely simulate and predict how the winning rate will change when players change certain tactics,” Wu said.
The Tokyo Olympic Games has officially come to a close. Chinese national table tennis team has once again proved its prowess. Like Zhejiang University sports big data innovation team, many other teams have lent their wholehearted support to Chinese athletes.