The Innovation Behind the TeamNL Game: Pose Recognition and Machine Learning

Tilburg, Thursday October 12th 2023 – You've probably noticed over the past 12 months how generative AI, a type of artificial intelligence that can generate content based on text, has skyrocketed in its advancements.

A crucial part of generative AI is training language models, a process commonly referred to as machine learning. There are various ways to engage in machine learning. In this blog, we'll discuss how we applied it in client projects, such as for NOC*NSF.

In today's digital age, capturing attention is a considerable challenge. To stand out, it's not just about informing your audience but also about entertaining them. When NOC*NSF presented us with a challenge, we saw an opportunity to blend both, using machine learning to offer their audience a unique experience.

The Diversity of TeamNL

TeamNL isn't just a collection of sports associations but a unified team representing a broad array of sports. NOCNSF's message is clear. The concept: an interactive quiz loaded with gamified elements with an innovative twist. The ultimate goal? To prominently feature this tool at physical stalls during NOCNSF events, like the European Cycling Championships or the Paris Olympic Games.

Game Concept

Imagine attending a NOC*NSF event and spotting a screen that invites you to "Start the TeamNL Game." You press start and see a character strike a pose typical of a sport, like baseball or volleyball. Your challenge? Mimic that pose. Succeed, and you advance to the next round, continuing until time runs out. Sounds simple, but behind the scenes, it's a technological marvel.

Technical Challenge

For the game to work effectively, accurately recognizing the player's posture is vital. The game employs a camera to capture the player's pose. This involves two significant challenges:

Real-time Recognition: The primary hurdle is tracking the player's posture in real-time. Thanks to a technique known as Pose Estimation, camera feeds can be converted into body key points. Among several options, Google Research's MoveNet model stood out for its speed and accuracy.

Determining the Correct Pose: Merely assigning fixed coordinates for a pose could lead to inaccuracies since everyone's physique is unique. Hence, we opted for machine learning. Utilizing Google's Keras API, a model was trained. After capturing and analyzing 17,000 images in various sports poses, we devised a unique model that can precisely determine if the player has assumed the correct pose.

Machine Learning: The Engine Behind Pose Recognition

Machine learning, a subset of artificial intelligence, allows computers to learn from and make decisions based on data without explicit programming. Essentially, it's a process where algorithms are trained to identify patterns and relationships in vast data sets. As the algorithm processes more data, it gets better at predictions and categorisations.

Our Application of Machine Learning

In developing the TeamNL Game, we used machine learning for sport pose recognition. In a specially designed studio, we photographed all desired poses. To get a comprehensive view of each pose, shots were taken from various fixed angles, ensuring every aspect of the pose was captured—vital for accurate pose recognition.

With an impressive dataset of 17,000 images, we began training our machine learning model. These images served as input for the model, enabling it to grasp the subtle differences and variations of each pose. The more images the model analysed, the more adept it became at precisely identifying a user's pose during the game.

The Impact of Machine Learning on the Game

Thanks to machine learning, the TeamNL Game can quickly and accurately assess whether a participant has accurately replicated a pose. Using such advanced systems ensures the game is fun, fair, and challenging, offering participants a seamless blend of technology and human movement.

In conclusion, by melding machine learning with an extensive sports pose dataset, we've designed a game that comprehends and appreciates human movement nuances. It's a splendid demonstration of how cutting-edge technology can enrich the sports experience and engage fans innovatively.

Standing Out in Experiential Marketing

In a saturated experiential marketing landscape, staying innovative and creative is paramount for relevance to the end-user. The TeamNL Game proves this is achievable when technology and creativity converge. It's a showcase of how brand experiences can transcend traditional methods to make a real impact.

Conclusion

In creating the TeamNL Game, we pushed boundaries and bridged technology and human interaction. For brands aiming to reach their audience in a deep and meaningful way, this project offers a glimpse of what's achievable when creativity and innovation go hand in hand. In an information-saturated and distraction-filled world, a well-crafted experience can indeed make all the difference.