Digitalisation and AI in amateur football

An article by Dr Lukas Breuer

How modern machine learning algorithms enable the live broadcast of amateur football

The Bundesliga, the German FA Cup, the Champions League and the Europa League – professional football’s illustrious competitions, known by every child and covered extensively by the media. Left in their shadow: The many amateur football clubs who take to Germany’s football pitches every week to play their sport with just as much passion as the professionals. Every match, every goal, all the big emotions and all the petty dramas unfortunately remain unseen except by the fans at pitchside.

That is changing right now. Start-up company soccerwatch.tv is using digital camera technology to record amateur football matches and broadcast them live, on demand or as highlights in order to increase their exposure. soccerwatch.tv is a start-up located in Essen, Germany, that had the idea of making amateur football in Germany accessible to the public. According to the German Football Association, 1.8 million football matches take place in Germany every year. 99 per cent of those are not broadcast. The reason? Previous solutions are cumbersome, expensive and substandard.

soccerwatch.tv developed their own camera system consisting of six Full HD cameras whose images are put together to create a 180-degree panorama view. This system is installed at a height of around 7.5 metres on the playing field’s floodlight tower in line with the halfway line and broadcasts to the world wirelessly in HD quality via LTE. In order to be able to deliver the image quality viewers of professional football are used to, an intelligent algorithm automatically identifies the relevant image sections from the panorama view in real time – without the need for a chip in the ball or the players’ shirts. soccerwatch.tv developed the algorithm for the artificial intelligence in this system in collaboration with adesso (acting as the IT service provider); adesso was so impressed by the idea that it invested in the start-up. But how do you develop these intelligent algorithms and how do you integrate them into your own business model?

From idea to first investor

One of the founders of soccerwatch.tv is the father of two football-mad children but often could not make it to their matches. This gave him the idea of installing a type of webcam at the football pitch, so he could follow the matches live. An initial prototype was made from a plastic box bought at a DIY store with a hole cut out of it and a camera installed inside, and the match was then uploaded to YouTube. soccerwatch.tv refined the idea to the extent that they won adesso as an investor.

“soccerwatch.tv’s solution is extremely innovative. That is why it was not even a question for adesso to financially and technologically support the start-up, for example by programming an intelligent algorithm for image recognition and a software program that ensures camera movements that the viewers’ eyes are used to, as well as by using the cloud,” said Christoph Junge, Member of the Executive Board of adesso.

Moulding an idea into an implementation concept

How do you turn an idea into an appropriate software concept? At the start of a data science project, it is necessary to develop a basic understanding of the business idea, the existing IT structure, the structure and quality of the data and what information the data contains. To ensure this understanding, a basic requirements analysis was carried out with soccer.tv as part of an Interaction Room workshop, a tried-and-trusted adesso method. The shared prioritisation and scopes enabled an appropriate implementation plan to be created and the right architecture and technology to be chosen. For soccerwatch.tv open source software combined with a cloud proved to be the perfect choice. Although as a young start-up soccerwatch.tv did not have the finances or staff needed to start with their own computer cluster immediately, a cloud system is exceptionally well-suited for their business idea. With every camera system installed the capacity can be scaled based on demand.

Proof of concept

In order to evaluate the functionality and feasibility of the concepts developed in the workshop, video footage was collected from a few cameras and used to develop the prototype of the intelligent algorithm as a first step. The development was fully agile. Agility in regard to data science projects has the advantage that the analysts are brought in very early on in the modelling process, unlike classic approaches where the data for analysts is only available right at the end. This approach allows prompt conclusions to be drawn as to whether the implementations deliver the desired results. This is even more important when an algorithm is the central component of the business idea.

From idea to launch

Once the feasibility had been demonstrated successfully, the algorithm was integrated into the cloud. The two essential requirements here were scalability and real-time capability. Weekend after weekend, 10,000 amateur football matches take place at the same time in Germany. Every camera system installed increases the processing load, which must be processed in parallel in order stream all these matches live.

soccerwatch.tv currently has around 65 camera systems installed, and the platform is in beta. As soon as the beta phase is successfully completed, the large-scale Germany-wide roll-out will begin – still supported by adesso, of course. So soon an uncle in Hamburg can follow his nephew’s matches in Munich live on his computer and then have a heated discussion with him afterwards.

More information at:soccerwatch.tv

Copyright: soccerwatch.tv