How AI will help keep time at the Tokyo Olympics

“In volleyball, we now use cameras equipped with computer vision technologies to track not only the athletes, but also the ball,” explains Alain Zobrist, Director of Omega Timing. “So it’s a combination where we use camera technology and artificial intelligence to do this. “

Omega Timing’s R&D department includes 180 engineers, and the development process started with in-house positioning systems and motion sensor systems, according to Zobrist, in 2012. The goal was to get to a point where , for several of the more than 500 sports events it works on each year, Omega could provide detailed live data on athlete performance. This data should also take less than a tenth of a second to be measured, processed and transmitted during events so that the information matches what viewers see on screen.

With beach volleyball, that meant taking this positioning and movement technology and training an AI to recognize a myriad of types of shots – from smashes and blocks to spikes and their variations – and types of passes, as well as the ball’s flight path, then combine that data with information gleaned from gyro sensors in players’ clothing. These motion sensors allow the system to know the direction of movement of the athletes, as well as the height of jumps, speed, etc. Once processed, all of this is then streamed live to broadcasters for use in on-screen commentary or graphics.

According to Zobrist, one of the hardest lessons for AI to learn was to accurately track the ball in play when the cameras could no longer see it. “Sometimes it’s covered by part of an athlete’s body. Sometimes it goes outside the frame of the TV, ”he says. “So the challenge was to follow the ball when you lost it. For the software to predict where the ball is going, then when it reappears, recalculates the gap from when it lost the object and picked it up, and fill it out [missing] data, then continue automatically. It was one of the biggest problems. “

It is this tracking of the ball that is crucial for the AI ​​to determine what is happening during the game. “When you can follow the ball, you will know where it was and when it has changed direction. And with the combination of the sensors on the athletes, the algorithm will then recognize the shot, ”Zobrist explains. “Whether it’s a block or a smash. You will know which team and which player it was. It is therefore this combination of the two technologies that allows us to be precise in the measurement of data.

Omega Timing claims its beach volleyball system is 99% accurate, thanks to the sensors and multiple cameras running at 250 frames per second. Toby Breckon, professor of computer vision and image processing at Durham University, however, is interested to see if this holds up during the Games and, more importantly, if the system is being fooled by differences in race and race. sex.

“What has been done is quite impressive. And you would need a big data set to train an AI on all the different movements, ”says Breckon. “But one of the things is precision. How often is he wrong in terms of these different movements? How often does he lose track of the ball? And also if it works evenly across all races and genders. Is it 99% accurate on, say, the US women’s team and 99% accuracy in the Ghanaian women’s team? “

Zobrist is confident and explains that while it may have been easier to bring in Google or IBM to provide the necessary AI expertise, it was not an option for Omega. “What is extremely important, whether it is for a scoring sport or a timing sport, is that we cannot have any discrepancies between the explanation of the performance and the end result,” he says. “So to protect the integrity of the result, we cannot rely on another company. We have to have the expertise to be able to explain the result and how the athletes got there.

When it comes to future schedule and tracking upgrades, Zobrist is low key, but says the Paris Games in 2024 will be key. “You will see a whole new set of innovations. Of course, it will remain around timing, scoring, and certainly also motion sensors and positioning systems. And certainly Los Angeles in 2028 as well. We have some really interesting projects there that we’ve just started.

More great WIRED stories

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *