Wednesday, 4 September 2024

Faster, fitter, cheaper: AI moves into sport

FEED

article here p128 - 135 

The sports broadcasting industry is poised to surpass $60 billion this year, a significant increase from $56 billion in 2023 and $54 billion in 2022. With so much at stake, broadcasters are innovating to attract new viewers and engage more closely with existing fans.

A new study from IBM, “Global Sports Attitudes and Behaviours,” reveals a growing generational shift and acceptance of technology-driven experiences that will impact the future of sports consumption.

It argues that by embracing digital transformation, adopting AI and data analytics, enhancing social media engagement, and supporting multi-device viewing, broadcasters can stay ahead of the curve and meet the evolving needs of sports fans.

A primary area of innovation is the integration of advanced technologies into sports broadcasting. Augmented Reality (AR) and Virtual Reality (VR) are revolutionising the viewing experience by offering immersive and interactive content that brings fans closer to the action in ways previously unimaginable. Artificial Intelligence and Machine Learning are also playing a critical role. These technologies are not only enhancing the personalisation of content but also enriching viewer experiences with real-time statistics and predictive insights.

IBM’s study shows that younger sports fans are more likely to favour AI-enhanced features as they increasingly turn to digital platforms for sports content. Nearly two-thirds (58%) of 18-29 year-olds polled believe AI will have a positive impact on sports.

Automated content generation is another key enhancement that can ensure timely and relevant updates for fans and help broadcasters keep their content fresh and relevant. AI tools can automatically generate highlights, summaries and other content, ensuring fans have access to the latest updates without delay.

In fact, AI in the sports market was valued at $1.4 billion in 2020 by IDC and is projected to reach $19.2 billion by 2030.

FEED spotlights the fresh developments.

Olympics

AI was extensively by Olympic Broadcasting Services during Paris 2024 to improve internal workflows, enhance the viewer experience, enrich storytelling and better explain the action.

It would not have been possible to produce 11,000+ hours of content, equivalent to 450 days of content compressed into just 18 days, without it.

Automated highlights were produced for 14 sports in Paris. Many of the algorithms on more niche sports like climbing having been trained by OBS from previous Olympic events.

Another AI tool was used to assist editorial teams to quickly generate highlights and to format them for different media including vertical video for social feeds. These AI systems pulled data live from commentary, auto-tagged the video and created automated summaries.

“We do not allow AI systems to auto-publish stories but it helps us a lot in identifying all the elements that make up good stories for posting to social media,” said OBS chief Yiannis Exarchos.

AI-based motion tracking technology also helped commentators and viewers keep track of athletes’ positions during events like canoe sprint, marathon and triathlon.

 

Automated content production

Greater visibility can make all the difference to club or league finances particularly in niche or lower tier sports where rights values are low if they are sold at all. Here, automated production systems are proving their worth. Pixellot’s system relies on PTZ cameras programmed to follow the on-field action using AI tracking. It automatically generates clips, for example of players who have scored, with functions processed in the cloud.   

According to the company, more than 150,000 games are broadcast every month in around 80 countries across 19 sports. This data feeds and improves its AI to increase the quality of productions.

A rival solution from Dutch developer Studio Automated has covered over 100,000 sports matches with installations in 1500 locations around the world. It is being used by Riedel Communications to build an AI-assisted video production solution. By joining forces, the companies aim to create a solution that will enable sports productions and leagues to remotely produce live broadcasts with minimal personnel and operating costs.

Paul Valk, Founder and Director of Studio Automated claims, “the maturity of our AI model enables us to serve clients that seek to automate production in the higher end of sports broadcasting.”

High-quality slow motion replays require expensive specialist camera systems – until now. EVS has developed the AI-based XtraMotion that allows production teams to transform any live or recorded footage into super slow-motion replays. The software drives other effects like deblurring, auto-cropping, and shallow depth of field.

Ross Video has added AI facial tracking to some of its studio cameras for live sports presentation. The Vision[Ai]ry range uses video analytics to automate the functions of a camera operator by using facial and body detection to locate and track the position of faces and bodies within the video stream directly from the camera. 

When using a facial tracking system, the AI can be used to accurately keep the talent in a specific frame and in focus. This frees up the camera operator’s time to prepare and frame the next shot.

AI-powered auto-tracking is becoming standard in PTZ cameras. Presenter Lock, a new technology from PTZOptics, allows its cameras to lock onto a specific subject and track even in crowded areas such as a sporting event or on-stage production. This technology leverages AI and computer vision, instead of simple motion tracking used in previous generations of auto-tracking cameras. 

Coaching and training

Even as the Paris Olympics ends us, sports federations are already turning to technology to prepare the next generation of Olympic champions for the 2028 Games in LA. Pixellot claims that its technology is transforming young athlete development at a scale unfathomable only years ago.

“By eliminating the need for manual filming, breaking down games and creating highlights Pixellot allows coaches to spend time on what truly matters: guiding athletes to excel,” says Yossi Tarablus, Associate VP Global Marketing.

Pixellot’s AI-automated end-to-end solution provides coaches with “a treasure trove of video and data,” enabling, for example, identification of strengths and weaknesses, and the creation of personalised game and training analysis.

“Our technology ensures that young talent from all backgrounds can be discovered,” says Tarablus “The platform’s video and data-driven approach helps federations and leagues refine their strategies by sharing and analysing game footage to uncover patterns and the stars of tomorrow. This level of insight is invaluable in the high-stakes environment of team sports.”

AI is gaining ground to plan analyse soccer training. Around ten percent of professional football clubs use data scouting at the moment, a number that will explode in the coming years. “At some point, all leagues will have this data,” Daniel Memmert, adviser to the German Football Association and the German Football League, told sports body ISPO. “You will then be able to click on a list of players with certain metrics that the respective club considers important.”

As sensor data from the ball and players is now highly reliable, everything is tracked accurately.  Tactical variables such as spatial control of individual players or entire team and pressing values can be analysed in fractions of a second.

Piracy

Piracy is an ever present threat with both sides taking AI to the battle. Cybercriminals are using AI to create highlights clips of sports which are then posted on TikTok to lure people into their illegal services.

“Pirates run sophisticated marketing organisations and AI is making their fake content look as good as if it were created by someone like the BBC,” says Tim Pearson, VP global solutions partner marketing, Nagra. “For example, what would have taken quite complex video editing to create a montage of the Olympic 100m final featuring previous winners like Usain Bolt, can now be done relatively quickly with AI tools. It’s all done as a shameless promotion of an illicit service.”

He adds, “AI gives pirates lots of promotion benefits that they use to create better clips driving traffic away from legal services. When content look this professional it dupes the consumer into thinking that the site must be legitimate.”

Nagra is using AI to fight back. “AI helps us find illicit shared content and confirm that it’s our customer’s property. Previously we scanned for patterns of activity and AI allows us to search much faster and for smaller patterns which may be focussed in a particular location.

“We also able to apply smart enforcement with AI. If AI detects an infringement, perhaps by identifying a watermark, then the model now has enough intelligence that it can disrupt that pirate by instigating counter measures.

In July the 2024-25 football season was a month away but Nagra is already detecting piracy and instigating proactive measures.

“We are detecting as much if not more piracy in this summer of sport.  There is a massive amount of content ripe of picking but AI is incredibly powerful for use in analytics, discovery, quantification and then remediation.”

 

Voice commentary  

With AI, sports teams can connect with their diverse fan base more effectively.  Platforms like Veritone Voice synthetic voice AI technology facilitate real-time commentary and descriptions in various languages. A partnership between Veritone and Stats Perform’s Opta sports data, offers not only lifelike voice options in a wide range of languages but voices like that of former pro football turned pundit Alan Smith. He has had his voice fed into Opta Voice and can now virtually commentate on soccer games in multiple languages.

Similar solutions are provided by Elevenlabs which is able to translate audio clips or the sound of videos into 29 languages to efficiently scale the localisation of content.

However, in the rush to expand reach, rights holders are wary about introducing AI too fast too soon, particularly if the audience is not ready.

Wimbledon shelved the use of AI commentary for this year’s championships following criticism of the “emotionless” feature on debut in 2023.

The model was trained to ‘read’ tennis matches and was only used for highlights commentary on Wimbledon’s website for matches on the outside courts where live human commentary was not in place. It was criticised by broadcaster Annabel Croft among others, for being robot-like.

The AI-powered audio, provided by Wimbledon’s data partners IBM, will continue to be used at the US Open, as well as golf’s Masters Tournament but may only reappear at the All England Club if the technology improves.

That said, Wimbledon and AI partner IBM continued to use AI to deliver a new mobile app that was intended to boost editorial coverage on the website. ‘Catch Me Up’ was built using IBM’s Granite LLM to provide AI-generated text and trained to mimic the Wimbledon editorial style. The Club also used Gen-AI to serve coverage of a broader range of matches than ever before, including wheelchair events.

 

Monetising the archive  

Advances in AI-powered digital asset management offer numerous opportunities for sports organisations to capitalise on their content. AI platforms make this possible via metadata tagging, content management, and e-commerce capabilities.  

“Past broadcasts and events hold significant value,” says Gary Warech, head of sports and entertainment, Veritone. “Through GenAI, archives can be rejuvenated and showcased innovatively through curated highlight reels, nostalgic montages, or other audio and video content types. This approach revitalises past content and helps carve out avenues for sustained revenue generation through sponsorships or licensing opportunities.”

Sports organisations can gain insights into their content through transcribed audio, logos, facial recognition, and other AI cognitive engines. NFL team San Francisco Giants described this speed and visibility into its archive with AI content management as having their own “personalised YouTube.”

AI is employed in MAM systems nearer production, for instance in EVS’ MediaCeption Signature solution. Its ‘natural language’ search capabilities enable non-technical users to more easily find what they are looking for.

“With AI, the potential to cater to diverse market segments—ranging from editorial teams and advertisers to the direct fan base—is exponentially magnified,” says Warech. “But it’s more than just a direct fan engagement tool. It serves as a means to do more with existing and future content, creating new content forms while extracting greater value from an archive.”

AI ad tools

New AI-powered content recommendation systems and advertising tools are opening more revenue streams. For example, AI/ML models dynamically identify the best spots on screen to deliver ads. “This is hugely important as viewers don’t want to be diverted away from the main event and forced to watch full screen ads,” says Nitin Jain, Managing Partner at Skandha Media Services. The company developed the Evince Ad verification tool for one of the largest OTT providers in India that currently owns the rights to stream a marquee cricket tournament in India.

“For live sports streams that may be firing hundreds and even thousands of ads over the full length of a live event, providing consistent and accurate data has been challenging,” Jain says. “The human resources needed to accomplish this level of verification just isn’t viable. We introduced machine learning to automate the process – training cameras to identify and capture ad elements. We can deliver super-efficient ad verification a proof of delivery that simply did not exist before.”

The Future of AI in sports

Metahumans or digital doubles might change the face of sports broadcasts.  During the 2022 FIFA World Cup animated replays created in Unreal Engine from official OptaStats match data showed new angles of play that were not captured by broadcast cameras. The AI-powered results, displayed on TUDN, the highlights show of Mexican broadcaster TelevisaUnivision proved popular. SMPTE suggested that the gamification of the event made it more accessible for those who normally wouldn’t be into the sport.

SMPTE also noted, in a recent paper, that the technique is expensive which perhaps explains why it hasn’t caught on. However, SMPTE suggests, in future people could rewatch entire games of their favourite sports from angles they have never imagined.

 

No comments:

Post a Comment