White Paper: AI Pre-Processing at the Edge is Transforming Sports Production

According to Allied Market Research, the global artificial intelligence in sports market is projected to reach $19.2 billion by 2030, growing at a CAGR of 30.3% from 2021 to 2030. With the rapid advancement of AI technologies, media and entertainment companies are increasingly leveraging AI-driven solutions to optimize their operations, improve audience engagement, and streamline content creation and distribution workflows. For sports, in particular, AI is being used to enhance and speed up archive search as well as improve post-game analysis, in-game activity, the fan experience, and more.

This article will explore the challenges of cloud AI pre-processing, along with the benefits and use cases for edge media pre-processing of sports content.

Challenges With Video Archiving and Cloud AI Pre-Processing

Jason Perr, AI Project Consultant at Perifery, a division of DataCore

One of the key challenges in sports content production is the need for faster turnaround times. The ability to swiftly locate and access valuable clips relevant to a sports event is crucial for users.

Historically, users have been at the mercy of manual tagging. If content isn’t tagged, or tagged accurately, it makes it near-impossible to find the most important and noteworthy clips. Looking for the right piece of content for a highlights reel could take hours – critically slowing down sports production.

Many sports media production organizations have adopted cloud-based media asset management systems along with AI cloud services like Google Vision or Google Video AI or Amazon Rekognition to process their vast and increasingly expanding library of valuable digital assets. These off-the-shelf tools  can be used to quickly add tags to content without the manual work of users. However, the quality of these fully automated tagging utilities varies drastically. Without being trained on specific data sets, and without having context into the specific kind of content being generated, the AI produces tags with a wide range of accuracy, making them untrustworthy. Added to the issues around consistent quality, the expense incurred to generate these tags, while being low (i.e., average of about $ 0.14 per minute for video tagging) can be hard to justify at scale. For a single NFL game, more than 100 hours of video can be generated when considering it’s a three to four hour broadcast with 25 to 30 cameras, along with pre-game and post-game content. The cost of generating tags for an NFL game is about $840 for tags that might not be that helpful. Due to this, many companies are trying more tailored AI models, trained specifically on the content they are working with. When an AI model is trained on a data set, it can make intelligent decisions, empowering sports organizations to find content quickly, without relying on the use of tags. However, in this case especially, the cloud has an unpredictable cost model and it requires significant time and effort to upload, download, and manage pre-processing services coming from various sources.

With live sports, the quantity of data being generated is massive since today it’s often being shot in 4K. Pre-processing services with AI have typically only been available in the public cloud. A major drawback of pre-processing AI in the cloud is that it’s expensive. Moreover, trying to move data back and forth between cloud and local systems, and shuffling metadata around, poses a security risk.

Benefits of AI Pre-Processing at the Edge

Running tasks at the edge vs. in the cloud eliminates the need for extensive data transfers, resulting in improved efficiency, cost reduction, and faster delivery. Performing pre-processing tasks at the edge is a more cost-effective and predictable solution compared to traditional cloud-based approaches. It allows sports organizations, regardless of their size, to optimize operational expenses while maintaining high-quality content production standards.

AI processing accelerates time of delivery by enabling sports organizations to perform critical pre-processing tasks on-site or remotely. Whether it’s analyzing player performance, identifying key moments, or generating highlight reels, AI at the edge allows sports organizations to quickly extract valuable insights and deliver engaging content to fans in near real time. By speeding up production processes, AI at the edge also enables content creators to focus on their craft, and frees up their time to concentrate on more creative tasks, which can lead to generating better, more engaging content.

Another benefit of AI pre-processing at the edge is improved monetization of archived content. Many sports organizations own large libraries of historical content, often with little assigned metadata. Generally, the older the content is, the less metadata. While the prevailing thought has always been to tag this content as thoroughly as possible into some kind of database or media asset management system, the problem is that when it comes time to search for something, the search is only as powerful as the quality of tags that were entered. In the best case, newer content can be found a bit more easily, and older content tends to get lost. In the worst case scenario, even the newest content is not easily found due to an incompatibility between the mindset or process of the person tagging and the person searching. With AI, it is possible to have a well-trained model review all of the content in a library and generate relevant tags, transcriptions, and summaries based on the content itself, logos, or even products on screen, and have this metadata associated with an asset record in a database or media asset management system. This information can provide valuable insight for working with advertisers and generating reports based on the quantity of certain kinds of content found within the library.

Even more interesting is the ability to train an AI to understand the content library. This is the future of media asset management. When an AI utility can understand the content in a library and comprehend what parts of that content are valuable and why, it’s quite valuable. Then if a user needs to access this content, they can simply request what they are looking for in natural language and get a result set that will almost always be more accurate and comprehensive than what’s possible through traditional tagging and searches alone. This type of advanced search feature would increase sports organizations’ ability to monetize content, enabling faster, more specific, and accurate identification of data.

In addition, AI pre-processing at the edge offers tremendous value to the eSports industry. eSports games, especially tournaments, can last many hours or even days compared with only a few hours for live sports, and can include thousands of players. Being able to automatically tag and search through a treasure trove of data is a real boon for the eSports industry.

Key Applications for AI-Powered Edge Processing

Some of the primary use cases for AI edge pre-processing include object recognition and facial recognition, which enables sports organizations to generate on-the-fly metadata effortlessly. With advanced object recognition, sports organizations can automatically identify and tag various elements within their content. This includes recognizing players, coaches, and referees, as well as detecting specific actions such as goals, fouls, or celebrations.

Object recognition also simplifies logo placements. With object recognition, sports organizations can ensure that team logos are displayed on basketball backboards or other key locations in an arena.

Facial recognition is another AI technique that can be used to identify players. By combining facial and object recognition sports organizations can identify players with greater accuracy.

Conclusion

The ability to deliver timely and compelling sports content is increasingly becoming crucial in the highly competitive sports media landscape. Today’s sports organizations need advanced AI tools and capabilities to stay ahead of the game.

Perifery is leading the charge for innovation with its new Perifery AI+ application-centric services for content production workflows. Perifery AI+ enables sports organizations to optimize their workflows, reduce costs, accelerate time of delivery, and unlock new monetization opportunities. With its seamless integration and advanced AI capabilities, solutions like Perifery AI+ are reshaping the way sports content is produced and consumed, ensuring fans around the world can enjoy the excitement and passion of their favorite sports in a more immersive and engaging way.