Spiideo AI game tracking system – automatically analyses and follows the play.

Fredrik AdemarSports, Technology

One of the most powerful features in the Spiideo system is the Virtual Panorama, that allows users to independently watch any detail of the game regardless of where it happens on the field.

However sometimes, e.g. when watching long game sequences or broadcasting the whole game, it is convenient to have the system automatically track the play and essentially act as a “virtual camera man” while you watch. This is exacly what the new Spiideo AI game tracking system helps you with, providing a very easy and fast way of viewing the game in such a way that is suitable for a coach or analyst. Below is a short example of this looks for e.g. football:

At any time you can take over the camera control, and navigate freely in the content (i.e. pan tilt zoom as usual in the virtual panorama). The automatic tracking of the game works from any angle. You can use the functionality on any device, or share it on a big screen.

Spiideo AI game tracking – how does this work?

The system relies on a combination of computer vision and Spiideo’s deep learning platform (i.e. a simulated neural network which in essence is a mathematical model based on how biological neuron networks works, such as the human brain).

Neural networks have received a lot of attention in recent years, because of their ability to solve complex and previously intractable problems. In this particular case the complexity lies in determining a pattern for where an individual, typically a coach (which is different from a normal spectator) would be looking at a football field during a match or practice. This means you need to e.g. analyse the overall game pattern and team structures (i.e. it is not enough to, for instance, only target thing like ball position).

To do this the system in short uses a combination of team/player detection (movements, positions etc) and a self learning neural network trained using various sports domain experts, while relying on large amounts of data for different sports (also taking advantage of input sources like e.g. eye movements using various eyetracking technologies).