Demo of a non-trivial application of video analytics, to compute automated ball possession statistics by team in a soccer (or football) match. We are using a sample match of Chelsea vs Manchester City.
It's built with a YOLOv5 detector (players & ball) and Norfair, our open-source multi-object tracking library for tracking objects and drawing their paths.
The possession is computed by counting the total time that each team has the ball. A counter is initialized when a team first has the ball and resets when the ball is taken by the opposite team.
Tryolabs is an AI & Machine Learning consulting company that provides critical expertise to transform business by building and deploying AI-powered solutions. We see ourselves as an AI-transformation partner, helping companies navigate during all the stages of the journey.
For more information:
- Blog post: tryolabs.com/blog/2022/10/17/...
- Source code: github.com/tryolabs/soccer-vi...
- Tracking library: github.com/tryolabs/norfair