1000 European football clubs lined up for Danish AI camera – University of Copenhagen

Faculty of SCIENCE > Press > News > 2018 > 1000 European football...

06 August 2018

1000 European football clubs lined up for Danish AI camera

ARTIFICIAL INTELLIGENCE:

With the help of the University of Copenhagen’s Department of Computer Science, VEO is now producing the world’s first intelligent football camera. The camera uses artificial intelligence to automatically follow a ball around a football pitch. Several Danish teams are already using it and more than 1000 European clubs are on the waiting list.

As of now, 16 Danish teams have purchased the camera, at roughly 10,000 DKK a piece, and more than 1000 European clubs are waiting to buy their own when deliveries begin this autumn. Foto: VEO

It looks like a well-designed green lunchbox with two eyes. Mounted on a four meter high tripod, the camera scouts across the pitch. The intelligent camera’s unique panoramic vision and high resolution allows it to film the entire football pitch at the same time. The camera has learned to track both ball and players as a game changes pace and location. As such, no cameraman is needed to record matches.

"Using 'Deep Learning’, the computer within the camera learns to recognize the ball and players on a pitch, their running patterns and how the ball moves." It is thereby able to predict a game's movements and turn accordingly," explains Postdoc Oswin Krause, of the Department of Computer Science’s Image Section, that works with image recognition in artificial intelligence.

1.5 million clicks feed artificial intelligence

To train the camera to recognize important aspects of football matches, it was necessary to get the right people to manually click on the ball and players. Currently, the camera intelligence is based on 1.5 million human clicks that have been inserted into a coordinate system of the pitch, as the computer continuously fills with different ball and player placements which it constantly learns from.

"Initially, it was difficult for the computer to discern a ball from other round objects such as shoe tips, knees or stadium lighting," explains Krause, who worked on the project for half a year.

The computer, or brain, in the camera is called a neural network. It tries to mimic the human nervous system with a series of small receptors that receive information and respond with an output signal. When the computer sees the ball move, it calculates where it is most likely to end up, and can thus move in the right direction at almost the same pace as the game. 

An analytical tool for coaches

The camera is made by VEO, a Danish company, and is on the market. It is now being used by several clubs, including Brøndby IF, Lyngby, KB and several amateur clubs. The camera’s ability to cover the entire pitch and the viewer’s ability to control camera angles after the match has been recorded allows coaches to follow certain areas or players on the pitch for a variety of thorough analyses. 

“Coaches can go in and see exactly what went wrong when their team loses a game,” explains Jesper Taxbøl - Chief Technology Officer st VEO.

As of now, 16 Danish teams have purchased a camera, at approximately 10,000 DKK a piece, and more than 1000 European teams are on the waiting list to buy their own when deliveries begin this autumn.

“No other product resembles this and clubs are interested, because it is hard for them to find cameramen to film for two hours," says Taxbøl.

According to Taxbøl, it is relatively easy to transfer the technology to other sports, such as basketball.