Metadata and AI – A Potential “Swiss Army Knife” for Sports Production

Like in film and television, sports fans are at least conceptually aware that when they watch sports, they are seeing a “finished product.” This product is the result of the work of numerous teams of production professionals all performing incredible technological feats.

But unless you’ve experienced (or are a part of) the “behind-the-scenes” world of sports broadcast, the names in the credits don’t hold the same importance as the action. This is odd, because the realities of sports production, whether broadcast live or not, require preparation, speed, strength, and skill rivaling that of the players.

Ok, that may be a bit of a stretch, but we know media production is very hard work – and you could argue that goes extra for the exceedingly high expectations and immediacy that come with sports. So as more and more sports garner attention, coverage, and production – there’s a lot to unpack.

So much input, not much output

The amount of footage created in a typical sports broadcast is insane. And depending on the perceived importance of the game, and the increase in scale it comes with, that insanity is compounded by orders of magnitude. One recent example comes to mind – At Super Bowl 50, CBS used a suite of 70 cameras to capture those fleeting moments of action between commercials – and that was two years ago.

Truth is, there’s an imbalance between what’s seen and good footage that never gets used. Continuing with the 70-camera example, for every shining moment or great shot, that means there are 69 equally-great-but-not-good-enough-for-prime-time shots left on the proverbial cutting room floor.

After a package rolls using the “best” clips, the remainder is often one massive bin, or if you’re really lucky, several bins (one per quarter/period/etc.) of disorganized footage, which seems like a huge waste. There are huge opportunities with this “leftover” content – whether as player features, sizzle reels, archival footage for licensing purposes, or just as a well-organized archive.

“The Big Guys” – like broadcasters, leagues, and rights-holders – typically have professional media loggers, interns, or short-term project hires to ensure everything gets logged and tagged. But remember, these are companies with deep pockets and even bigger IP recognition.

It’s exactly the type of exhausting, tedious work that only big companies have the clout to entice someone to do, where “willingness to work overtime, on short deadlines, and on long weekends with short notices” and things like, “…[anyone] whose logs are sloppy will not be asked to extend” are common in job descriptions.

Even then, after all of the work is done, most of it will still end up as a logged, organized bin of clips sent to archive. And this is in the “big leagues.”

The practical outcome is that if you’re producing sports at anything but the highest level, unused footage might as well still be in the camera.

Metadata to the rescue…ish.

Metadata is data that references other data. Though it was available long before the switch to digital, it had rarely been used except as the byproduct of using digital equipment – now it’s becoming a critical component of sports production and media production in general.

Dr. Richard Chernock, Triveni Digital‘s chief science officer and chair of ATSC’s Technology and Standards Group, boiled-down metadata’s new importance to digital broadcast in a 2017 interview with IBC.

“Metadata is the key to making all of this work together,” says Dr. Chernock. “As one simple example, without signaling metadata, a receiver would not know which streams to decode in unison to provide the video and audio of a TV program, so the metadata for signaling and announcement is a core part of making linear TV work, and will continue to be that into the future,” he says.

He also mentions that television has outgrown its linear model in certain instances – like sports – which adds another level of importance for metadata management.

Television, production, broadcast, multi-streaming, and the habits of viewers continue to enrich traditional programming with content created from multiple sources. It’s not surprising that metadata has also increased in its importance.

Where AI and metadata merge

So, metadata is here to save the day for the fast-paced world of sports media production, right? There are multiple media asset management options on the market that should help production staff get a handle on their metadata and not have to worry about anything.

The problem: Many MAMs rely heavily on an editor or logger manually editing metadata based on subjective analysis of the clips. While things like the score and the timing of penalties can be synced to footage timecode, there’s still a glut of potential content out there.

Metadata, then, is less “skeleton key” and more “lock-picking kit” for much of the sports production use case, with human input needed for anything that falls outside of objective data. However, there’s an interesting case emerging for the use of artificial intelligence to bridge that gap.

AI has recently been a buzzword to the point that it’s already starting to lose its meaning. AI this, AI that…AI can detect a hit from a Wiffle ball bat.

It’s true, the industry has been working hard to bridge the gap between what AI can achieve and human production work, because not everyone has the budget for professional loggers. As AI matures, its value will continue to increase as it learns what’s important in each sport.

AI poised for a spot on the top line

The NHL’s Vegas Golden Knights enjoyed a whirlwind first season – falling just three wins short of bringing the Stanley Cup home to T-Mobile Arena. While both the product on the ice and the spectacular opening pre-show for the playoffs received plenty of attention, Andrew Abrams, the Golden Knights’ Director of Post Production, knows there’s more work that could be done – if only there were a way to add more time to a day.

“During the regular season, we have a day – maybe two – to get game footage turned around before we’re back at it for another game, and the playoffs become an even bigger time crunch,” says Abrams, “Even with our whole team, and sometimes even when an extra staff member can dedicate about half a day to logging, there’s just not enough time to organize anything as much as we’d like.”

On top of that, hockey itself is a fast-paced sport. The constant action, few stoppages, and on-the-fly line changes represent a high degree of complexity when logging clips. As AI matures, it will eventually reduce the burden on production teams by handling things like transcription, translation, facial recognition, product recognition, and much more.

For everyone from the team in the truck, to the team in the broadcast booth, to teams like Abrams’, the prospect of AI being able to integrate that type metadata into their workflow through AI is an attractive one.

“We’re creative people, and we hire people for what they can bring to our production, not for how well they manage assets,” says Abrams. “If we could be more of a final check on what AI sees in the metadata – whether it sees two bodies collide and tags that as a hit, or even something like a player’s numbers – and confirm it’s tagged correctly before using it, that would free up our minds to keep creating what fans want to see most.”

In the near future, organized metadata, with a little (OK, a lot of) help from AI, may just form the underpinning of future successful sports production – and all because of that “background data” which had always been available.