| General Info | |
|---|---|
| Working Title | CompVis FTRS |
| Student | Joshua O'Donnell, jodonne5@tcd.ie, 25353241 |
| Student | Viraj Bhor, bhorv@tcd.ie, 25358463 |
| Start Date | 31.10.2025 |
| Study Program | Computer Science M.Sc. Augmented & Virtual Reality |
This Repository will contain our project for the Computer Vision module in our M.Sc. program at Trinity College Dublin.
The project consists of implementing any application of Computer Vision, we have chosen to implement an incrementally extendable tactical recognition program for football/soccer matches. The project will proceed as follows:
- The Minimum Viable Product will recognize players and the ball's position on the pitch from a static screenshot of a football match.
- In further iterations, we will then expand into one of the following features:
- Recognition in motion - tracking players and already implemented features on video input, rather than static imaging,
- Formation matching - using a provided base formation, draw lines between the players to mirror their starting lineup,
- Space recognition - detect open spaces the attacking team can potentially use to progress the ball,
- Passing lane recognition - suggest safe, progressive passing routes for the player in control of the ball,
- Passive feature tracking - detect information such as the direction players are looking in, whether they are on a run, or standing more or less still, or, using video input, a progressing run with 'history', player speed, ball speed etc.
The final product will be due for delivery by TBA, and defended in a small presentation in December or mid-January, after which the module is closed for this academic year.
For time constraints, Anthropic's Claude provided a starting point for the implementation of the "self-made" model. The exact code pasted from Claude is visible in ftrs.py at commit "Init ftrs.py and Python 3.12 env (again)" (https://github.com/JoshDe777/CompVis_FTRS/commit/7b3b2150dd6182f28fdc932c37d917468e3c9381). This code was inspected and adapted by the project participants in subsequent commits to adjust the model's capabilities and contents to our effect.