Credit Image: Waterlo University.
Waterloo University researchers have developed a new artificial intelligence technology (AI) that can analyze the performance of the launcher and mechanics accurately using a low -precision video of baseball games.
The system, which was developed in Baltimore Oriles by the Waterloo team, connects holes with more detailed and expensive technology that has already been installed in most of the stadiums hosting the main baseball game (MLB), whose teams have increased in data analyzes in recent years.
Waterloo researchers convert the video of a jug to a two -dimensional model that AI in Pitchernet can later analyze. (Credit: Waterlo University)
Those systems, produced by a company called Innovations Hawk-EyeUse multiple private cameras in each park to catch players at work, but the data they produce are usually available to the host team that has the stadium that runs these games.
To add games away to their analyzes, as well as use the smartphone video taken by the scouts in small league games and university games, Oreols requested video experts and artificial intelligence experts in Waterloo to get help about three years ago.
The result is a relatively simple system called Pitchernet, which overcomes challenges such as non -clarity of movement to track the movements of jugs on the hill, then results in data on standards including the pitch and the version point of standard broadcasting and smartphone.
Researchers at Waterloo used images created during the training process to help build Pitchernet AI technology. (University of Waterlo)
Dr. John Zelik, Professor of Systems Design and Participated Directors of the Games Design, said, ” VIP vision and processing laboratory (VIP) In Waterllo.
“The goal of our project was to try to repeat the Hawk-Eye technology and bypass it by producing similar outputs from the broadcast video or the smartphone camera used by the scout that sits somewhere in the stands.”
To help train artificial intelligence algorithms at the heart of technology, researchers have created three -dimensional deities of jugs so that their movements can be seen from many preference points.
The broadcast video clip is used from the center’s field to create a three -dimensional human model by the Pitchernet system. (University of Waterlo)
Information from the video processed by the system is provided to the vital mechanics of Oriols, who committed to financing the project jointly for another year.
This data can be used to adjust how to throw ball jugs to improve performance or avoid injuries, evaluate future success and the durability of the pitch horizons.
“The current technology has already improved baseball analyzes,” said Jerene Bright, a doctor who had a pioneering role in the project. “Since it is limited to home games, there is a real need for solutions that work anywhere, especially for scout. This is where our system comes.”
Researchers are now exploring the basic idea application – the player’s artificial intelligence analysis that introduces the use of the standard video of smartphones and smartphones – on other professional sports, including hockey and basketball, as well as other aspects of baseball, such as beating.
A paper on the project, Pitchernet: Moneyball Evolution operation in baseball video analyzesIt was presented at the 2024 IEF/CVF conference on computer vision and patterns.
Waterlo University
