How An AI-Automated Sports Broadcaster Is Shaking Up Soccer Streaming

TV edit suites like this are not required with Pixellot's automated systems. (Photo by Peter Dench/Getty Images) GETTY IMAGES

Pixellot is the biggest sports broadcaster you’ve never heard of.

The Israeli company does not have the audience of an ESPN or DAZN, but in terms of the number of events it produces, it’s the world leader.

Using artificial intelligence (AI) and automation, the company produces around 100,000 hours of sports matches every month. Last year, Pixellot produced more than 220,000 live matches – from soccer and basketball to hockey and handball.

Previously, these events would have only been seen by the (often small) groups of supporters physically standing on the sidelines or sitting in the stands.

“We realized around 99% of all games in registered leagues are not streamed or professionally captured,” Pixellot CEO Alon Werber told me in an interview.

“We thought if we automate the entire process of producing, streaming and adding other options, there might be a very large market.”

Today Pixellot, which was founded in 2014, has its systems installed in about 6,500 venues around the world, with 40% in the U.S.

Once installed, Pixellot’s cameras provide a panoramic broadcast of a sporting event that can be live streamed. Different angles and additions like live stats are available, but a big reason for the company’s growth is its broadcasts do not require human camera operators or production staff.

“Even if you pay one person, it can become prohibitive. And if someone is using their smartphone, it doesn't look good, people can't clearly see the game, and it's not live,” Werber says.

“In most of our systems it's completely automatic. No one needs to do anything apart from scheduling the game and defining the type of sport. It is all done through a combination of computer vision, artificial intelligence and algorithms.

“In one U.S. high school that has our system they schedule a season in advance with tens of thousands of games and that's it. No one needs to do anything, it's all set automatically.”

Pixellot systems are installed in professional and amateur competitions. PIXELLOT

The AI element sees each of the cameras in one of Pixellot’s systems covering a slice of a pitch, field or court where the game is being played. Each slice is then “stitched” live into one panoramic view.

“We then follow all the moving objects all the time and we detect the state of the game,” Werber says.

“So we understand based on the computer vision and algorithm that this is a corner (in soccer) and not just two players near the corner, for example. If it's a corner the camera needs to behave differently and now it's a virtual camera, it's a virtual frame within the entire panoramic view. What dictates a virtual frame is the rules and the AI.

“We have a database of more than half a million games, 200,000 of them soccer. As we have more and more games, the AI improves accuracy. We apply deep-learning technologies to track the ball better and we provide better and better accuracy of the framing.”

The cost-effectiveness of such an operation has seen numerous leagues and competitions that were previously not broadcast, particularly in soccer, sign up with Pixellot.

The company recently announced a deal to work with IMG ARENA to produce live coverage of the next season of the Scottish Championship, Scottish soccer’s second tier. While in Argentina, Pixellot is working with local broadcaster La Corte to show Primera A, the top division of women’s soccer. It is the first time the competition has had a broadcaster.

BUENOS AIRES, ARGENTINA - APRIL 6: Stephanie Melgarejo of River Plate during the 9th round match between River Plate and Racing Club as par of the Primera División Femenina 2018-19 at River Plate facilities on April 6, 2019 in Buenos Aires, Argentina. (Photo by Diego Haliasz/Getty Images) GETTY IMAGES

Semi-professional and even amateur leagues in everywhere from Germany to Mexico are also clients, as are U.S. colleges and high schools.

“Although the potential audiences are smaller, they are often very engaged,” Werber says.

“It might be parents (of high school athletes), it might be local people, it might be women’s sport, which we cover a lot.”

At the top end of the sport, Pixellot also works with some of the world’s biggest soccer clubs, including Real Madrid and Bayern Munich, to provide coaches and scouts with video analysis tools. Match clips can be cut and instantly shared or individual player highlights automatically generated, for example.

LISBON, PORTUGAL - SEPTEMBER 05: Alon Werber CEO of Pixellot speaks at the Academy room during Soccerex Europe - Day 1 at Tagus Park on September 5, 2019 in Lisbon, Portugal. (Photo by Carlos Rodrigues/Getty Images) GETTY IMAGES

Werber says the company’s revenue is “in the tens of millions of dollars”. He says the business will be profitable by the end of this year, due to investment in R&D including more applications for youth players.

“When you do those things you open new, large markets, but you have to invest quite heavily.”

While Pixellot also works with traditional broadcasters on more complex productions where humans are required, Werber thinks automated production will continue to grow.

“Of course, we are not trying to imitate the production experience in Barca against Real Madrid, but we want to have a very good production experience where you can see scoring, you can see graphics, you can zoom in and zoom out, and everything is automatic.

“I think AI-driven production will take more and more slices out of regular production. I'm not saying it will take tier one, very professional, 17-camera productions very soon, but it will increase.”

Form:forbes.com