Semi-automated offside: the AI that tracks 22 players
Semi-automated offside combines two things: cameras that track all the players and a sensor inside the ball that detects the exact moment of the pass.
The cameras feed a computer vision system, a type of artificial intelligence that recognises each player in the image and estimates the position of many parts of their body, several times per second. With that data, the system draws the lines and calculates whether, at the moment of the pass, an attacker was closer to the goal line than the second-last defender.
When it detects an offside position, it alerts the video referees, who review the play and confirm before it is communicated. It also generates a 3D animation shown to the crowd.
It is faster and more precise than drawing lines by hand, but it has limits: camera quality, players blocking each other from view and the need for a person to supervise every decision.