Automatic mapping of football team formation using computer vision
Αυτόματη αναγνώριση σχηματισμού ομάδας ποδοσφαίρου με μηχανική όραση
View/Open
Keywords
Υπολογιστική όρασηAbstract
The integration of artificial intelligence in the area of sports is showing an upward
trend the past few years. There is now a vast amount of generated data that can be
processed and transformed into useful information in various aspects of a high demand
sport such as football. In this thesis, primary target is the extraction and visualization
of tactical statistics from broadcast videos of football matches. Continuous video frames
are being processed and analyzed, in order to detect players positions and project both
teams formations on a static top view of a football field. The emerging knowledge from
the results can be valuable for the examination of the tactical team’s performance during
the game. Deep learning methods from the area of computer vision are applied to fulfill
the tasks of players identification and camera pose estimation. A state of the art data set
is used for the evaluation of different frameworks and the whole procedure is validated
using image sequences acquired from Greek football matches. Ground truth data were
generated manually for the two main tasks and the predicted results were compared with
them to measure their accuracy.
Postgraduate Studies Programme
Πληροφοριακά Συστήματα και ΥπηρεσίεςDepartment
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα Ψηφιακών ΣυστημάτωνNumber of pages
69Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα