Skip to content

tomkeane07/rugby_tracker

Repository files navigation

rugby_tracker

This was developed for a MSc thesis in Artificial Intelligence at the University of Limerick in 2022. This project received above a grade of ~95% and was awarded a 1st class honours overall.

Here's a copy of my thesis, describing what the application does, and how.

And here's a copy of a more concise presentation on the subject.

Rugby Tracker GIF

Abstract

Sport field registration (SFR) and projection to a template field map can provide unique insights to supplement sport analytics and coaching. SFR and projection have been utilized to various degrees across different sports, including football, American football, basketball, and ice-hockey. The challenges imposed on SFR differ in scope and qualitatively across and within sports due to variations in pitch dimensions, field markings, application of camera pan & zoom, among others. This thesis proposes a sport field registration model based on Canny edge detection and Hough line transformations of a rugby broadcast footage, and then contextual semantic rugby field line classification. Classified lines are used to provide a homography matrix that projects player positions to a template field map. Player positions are provided prior to the homography transformation by an Ultralytics YOLOv5 pedestrian detector which is run on each broadcast frame.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published