In this case, we created a mobile application that can be used for soccer training purposes. The AI in this app tracks various activities (like e.g. juggling and jumping jacks) that are executed by the person that is being filmed. The app uses multiple techniques, like ‘object detection’, ‘pose estimation’ and ‘tracking solutions’, in order to make all this work.
- 10.000 lines of code
- 8+ exercises
- 60.000+ images used
Although soccer is one of the world’s most popular sports, the technological advancements in the sector are lacking behind. A regular camera would most likely be an important tool in your toolkit, if only you were to play at one of the big clubs. Video shoots serve the purpose of validating whether the players did their homework. Currently, these videos end up in storage, never to be watched again. There is, however, tremendous value to be gathered from these videos. Both exercise completion, as well as skill measurements are easily made out. All you need is the relevant know-how to create this. Luckily that is something which is abundantly present at AI Heroes.
For such a project, the key to success is understanding the problem well. This required us to get a 360 scope of the task at hand. We’ve talked to various stakeholders, and came up with a preliminary design. That would be modular enough to handle the scope of the project.
Using our AI Heroes Iterative Approach we were able to get a juggling exercise ready within 12 weeks, and an extensible framework containing multiple exercises within 6 months.
The techniques we have used were (among others) a combination of object detection, pose estimation, and tracking solutions.
Of course, the techniques used for this assignment could be easily repurposed to completely fit your use case. Especially object detection is widely applicable, as it is a very versatile technique. Possible use cases could be: model detection, quality assurance, corrosion detection, dirt detection or every-day object detection.