AI Football Training
For our partner JOGO, we helped create a mobile app to improve the performance of football players. With the help of AI Heroes, JOGO’s AI in this app tracks and analyzes various exercises, such as juggling and jumping jacks. The AI analyzes players’ performance and provides data-driven insights to reach their full potential!
Technologies used: Computer Vision, Pose Estimation, Object Detection
Introduction
Football is by far the most popular sport in the world but AI in football is unexplored territory. There are some sports out there where data-driven insights mean the difference between winning and losing. Why not use AI to improve the performance of players on the football field? JOGO saw an opportunity to help players reach their full potential using AI technology. That is why they built a data-driven football platform, where players, data and science all come together. And to get the best data-driven platform you need to harness the power of AI!
Goals
Many players film their training on their phones, re-watch the videos to see how well they did and what they can improve on. JOGO saw that AI could help players and give them tips on how to improve their performance. What was needed was an AI that can understand what players are doing and how well they do it, with a computer vision approach! As players mainly used their phones to record what they did, the AI also needed to be so efficient that it could operate on edge devices with very limited computational power. To make this possible, JOGO and AI Heroes joined forces.
Computer vision is key
For JOGO, player-recorded training sessions are central to their platform. The question they gave to AI Heroes is: ”How AI can translate these videos into performance-improving insights”. The first step was getting from video to data, with computer vision. The app uses object detection to detect the relevant things in the video, such as the player and the ball. The AI also needs to know what the person is doing with the ball! That is where pose estimation comes into play. Pose estimation is a machine learning model used to detect in which pose a person is, and how a person moves.
The outcome
The AI Heroes’ Iterative Approach enables us to build prototypes and deliver proof of value very quickly. In JOGO’s case, our high-speed method meant that they could see if their idea would work in reality within 12 weeks. The AI that we built together recognizes players doing various exercises with and without their ball. Now JOGO is the leading AI-driven platform for football!
How can your organization benefit? Techniques like the one used here, can also be used for your needs. In health care, for example, we use computer vision to detect the safety of people. But you can also use computer vision to detect non-human objects. Think about tracking animal well-being or using it for quality assurance on assembly lines. With our ready-made building blocks we are able to realize your AI idea in the shortest possible time!
