Team: Sami Ibishi, Andac Kurun, Yuwen Lu, Jonathan Mengedoht, Marie Schwahn, Kerkko Visuri
Date: April–July 2019
My role: Design and front-end-development (Swift, iOS 13)
Goalplay, founded by Oliver Kahn and Moritz Mattes in 2016, provides goalkeeper training plans, content and gear for individuals and football clubs. The company is working on an app to digitalize goalkeeper coaching.
Most amateur goalkeepers don’t have a specialized coach. The football club coach spends most of the time with the field players. Goalplay wants to create a digital coaching app to provide goalkeeper coaching to aspiring players all around the world. They asked us to build a proof-of-concept app that records, classifies and analyzes a goalkeeper’s exercises. The app should also provide feedback after the training session. It should work on a mobile phone to be the goalkeeper’s everyday practice companion.
We worked with Goalplay and real goalkeepers to create an app with two main features: virtual coach and activity recognition engine. With the virtual coach, the user can create personal training plans based on their desired training frequency and workload.
During training sessions, the app records his/her training, while our activity recognition engine uses Machine Learning algorithm to classify the user’s performed exercises live and display the results on the screen. To create the activity recognition engine, we created a proprietary LSTM model which can classify common goalkeeper exercises with over 90% accuracy.
We delivered a working proof of concept app to Goalplay after three months of development. The app can classify common goalkeeper exercises with high accuracy, high FPS and low battery drain. This is enabled by using highly optimized CoreML model and ARKit for motion capture.
Based on the promising results, Goalplay is planning to further develop this technology.