Full width home advertisement

Welcome Home

Post Page Advertisement [Top]

Artificial intelligence has the potential to make ice hockey analysis significantly easier to understand


In professional sports, every decision is based on data and analytics. Sports science is a fast-paced field, and in order to be successful, every professional team must make use of it. However, when everyone on the field appears to be identical, such as in ice hockey, it can be difficult to conduct an analysis. Do you, on the other hand, know what can be of assistance? As discovered by researchers at the University of Waterloo, Artificial Intelligence is a type of artificial intelligence.


Waterloo is located in Canada, a country that is well-known for its involvement in the sport of ice hockey, among other things. And it's a visually stunning game that necessitates dexterity, passion, and incredible fortitude. Players, on the other hand, are extremely difficult to track down due to the fact that they all wear the same uniforms and helmets. It's made even more difficult by the fact that they're constantly changing without pausing the game. In light of the game's rapid-fire nature, this is particularly true. Even though there are some issues with the game, fans can still enjoy it despite them. However, analyzing the game is quite difficult and requires keen eyes.


Consequently, scientists amassed a collection of over 54,000 images from National Hockey League games as a result of their research. In fact, this was the largest data set of its kind ever assembled. They then used this information to train their algorithms to recognize the numbers on the sweaters of the players in question. This is a difficult task, but it is also critical because the number is the primary cue used to identify a specific player in an ice hockey game, so getting it right is essential. True, the face can be seen as well, but it is obscured by the helmet in this instance. However, it appeared that the artificial intelligence solution was successful, as the researchers achieved an impressive level of accuracy.


The scientists improved the accuracy of their AI by instructing it to treat two-digit numbers as if they were made up of two distinct digits added together. The accuracy of the algorithm was increased to nearly 90 percent as a result of this simple concept, despite its simplicity. "Using different representations to teach the same thing can result in improved performance," says Kanav Vats, the project's director. Our combined wholistic and digit-wise representations yielded excellent results," says the researcher.


Researchers are now working on artificial intelligence (AI) that will be capable of tracking players in video, locating them on the ice, and recognizing their actions in real time. A shot is being taken or an opposing player is being checked by the algorithm, which will be able to distinguish between the two.


The question you may have is why artificial intelligence is required for this type of task in the first place. The process of conducting a manual analysis is feasible, but time-consuming. Because their uniforms and helmets are nearly identical, it is extremely difficult to distinguish between them when watching videos. But it is required for the purposes of analysis of the game's tactics and for the purpose of illustrating certain points during the broadcast. Additionally, it demonstrates the practicality of artificial intelligence solutions in nearly every aspect of daily life.

No comments:

Post a Comment

Bottom Ad [Post Page]