Very good question! - Well, maybe...

The way we train our machines to recognize the ball and players is primarily based on over 1.5 million ball annotations. That means, we have fed a computer with a lot of images from a football pitch where we have pointed out where the ball is located. In that way we can train a computer to make good decisions on where the ball is at any given point.

For that to work with other sports, say basketball, we need to do the same work over again with images of basketballs.

At the moment we have focused on getting high quality ball detection in football, but we will start exploring other sports during 2018.Β 

Let us know what kind of sports you are looking for, and we might find a way to make it work.

Did this answer your question?