Deep Learning for Soccer Game Analysis

Master

Agadus is a full system ranging from capturing data from cameras to delivery of individual camera angles and sequences to end-users. This includes parallell and distributed processing on CPUs and GPUs, high-speed interconnection networks, computer vision and video management, usable user interfaces in browsers and mobile apps, etc. Thus, there are many opportunities for an interesting master topic, both for single students and groups.

Description of topic

In this master project, we aim to further develop BAGADUS which is currently in use at Ullevaal Stadium for the National soccer team and at Alfheim stadium for Tromsø IL.

Goal

The goal is to utilise deep learning for player and ball tracking in the Bagadus system.

Learning outcome

Image and video processing skills and machine learning.

Qualifications

Signal processing, good programming skills

Supervisors

  • Pål Halvorsen
  • Michael Riegler