The digitalization increasingly influences many top-class sports. Technological advances enable precise collection of positional data, which is, however, only of minor benefit in practice. The goal of this cooperation project is the development and visualization of sport-relevant game and tactical analyses. This is to be achieved through the intelligent multimodal fusion of video, position and event data. In combination with suitable expert models, this enables the provision of digital assistants for sports analysis.
The MM4SPA project is a transfer project from the field of sports analytics. Goal of the project is the prototypical integration of analysis methods from the area of machine learning into the existing analysis platform of the industrial cooperation partner KINEXON. This will lead to the development of new, innovative services in the area of sports analytics.
Based on available sensor systems from the cooperation partner for gathering positional data, an automated merging with video data for sports games is being developed. Through this multimodal fusion, the data is semantically enriched by the automatic detection of sport-specific events and their evaluation using collaboratively developed expert models. This enables the provision of digital assistants to support customer decision-making processes.
For this purpose, the cooperation partners use methods of video analysis and computer vision in combination with Machine Learning processes.
To ensure the flexible use of the methods, the various sports game disciplines football, basketball and handball are dealt with. The disciplines differ with regard to the complexity and dimensionality of the data and enable the development of a flexible and future-proof system. In addition to the ability of the industrial partner to use AI applications, close cooperation and the associated knowledge transfer ensure that the solution can also be independently developed by small and medium-sized companies in the future.