Cornell's Laboratory for Intelligent Systems and Controls developed an algorithm that can predict volleyball players' in-game behaviors. According to the research, it can detect the players' movements with 80% accuracy.
Volleyball Blocked Shot
Machine Learning for Predicting Players' Movement
The algorithms were trained by Ferrari and PhD candidates Junyi Dong and Qingze Huo to infer hidden variables in a manner similar to how people learn about sports through watching games. The algorithms used machine learning in order to extract information from volleyball game videos and utilize it to create predictions when presented with new games.
The results were published in the journal ACM Transactions on Intelligent Systems and Technology. It demonstrated how players' roles can be inferred by algorithms. For instance, the algorithm was able to separate a blocker from a defense-passer with an average accuracy of around 85%. On the other hand, the algorithm also predicts several actions across a series of up to 44 frames. In this case, it has an average accuracy of more than 80%. Spiking, setting, blocking, digging, jogging, crouching, falling, standing, and jumping were among the movements.
Holistic Algorithm in Predicting In-Game Behavior
According to TechXplore, the algorithms are unique in that they take a holistic approach to action anticipation, combining explicit information like an athlete's position on the team with implicit information like their location on the court.
The study's principal investigator Silvia Ferrari, a John Brancaccio Professor of Mechanical and Aerospace Engineering, stated that computer vision can decipher visual data such as jersey color and a player's position or body posture. She claimed that, while incorporating hidden elements like team strategy and player responsibilities, they continue to employ real-time information.
ALSO READ: The Science Behind the YouTube Algorithm
Benefits of Algorithm During Game Preparation
Ferrari hopes that teams will use the algorithms to practice specific moves and game scenarios before competitions by teaching them from video of an opponent's previous games and exploiting their predictive abilities.
Ferrari has submitted a patent application and is currently collaborating with the Big Red men's hockey team to advance the program. Ferrari and her graduate students, under the direction of Frank Kim, are developing algorithms that automatically recognize players, actions, and game settings using game footage provided by the team. One objective of the initiative is to assist in the painstaking manual annotation of game film by team staff members.
Ben Russell, director of hockey operations for the Cornell men's team, said that their program places a major emphasis on video analysis and data technology. He said that they are constantly looking for ways to evolve as a coaching staff in order to better serve their players. He added that he was very impressed with the research Professor Ferrari and her students have conducted thus far. He believes that this project has the potential to dramatically influence the way teams study and prepare for competition.
Ferrari said that, beyond sports, the capacity to predict human responses holds great promise for the future of human-machine interaction. He added that better software can enable autonomous vehicles to make better decisions, bring humans and robots closer together in warehouses, and even improve video games by boosting the artificial intelligence of the computer.
RELATED ARTICLE: New Anti-Crime AI Algorithm Predicts Activities One Week Ahead With 90% Accuracy, Reveals Bias in Police Work
Check out more news and information on Technology in Science Times.
