https://github.com/zakroum-hicham/football-analysis-cv

This repository contains a computer vision/machine learning football project that uses YOLO for object detection, Kmeans for pixel segmentation, and perspective transformation to analyze player movements in football videos
https://github.com/zakroum-hicham/football-analysis-cv

ai computer-vision data-science football-analytics kmeans-clustering machine-learning opencv yolov8

Last synced: 16 days ago
JSON representation

This repository contains a computer vision/machine learning football project that uses YOLO for object detection, Kmeans for pixel segmentation, and perspective transformation to analyze player movements in football videos

• Host: GitHub
• URL: https://github.com/zakroum-hicham/football-analysis-cv
• Owner: zakroum-hicham
• Created: 2024-10-15T00:01:16.000Z (about 1 month ago)
• Default Branch: main
• Last Pushed: 2024-10-24T21:41:57.000Z (21 days ago)
• Last Synced: 2024-10-26T09:27:20.477Z (20 days ago)
• Topics: ai, computer-vision, data-science, football-analytics, kmeans-clustering, machine-learning, opencv, yolov8
• Language: Jupyter Notebook
• Homepage:
• Size: 42.5 MB
• Stars: 13
• Watchers: 1
• Forks: 5
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# --> Football Analysis Project

##  Project Overview

The goal of this project is to detect and track players, referees, and footballs in a video using YOLO, a leading AI object detection model. I also trained the model to improve its performance. Additionally, we assigned players to teams based on the colors of their t-shirts using Kmeans for pixel segmentation and clustering. We also assigned the ball to a player during gameplay. With this information, we measured a team's ball acquisition percentage throughout the match. Furthermore, we implemented perspective transformation  to convert the scene into a 2D graphic,, allowing us to measure a player's movement in meters rather than pixels.









  





  



## Features

- Object Detection: Detect and track players, referees, and footballs using YOLO.

- Player-Team Assignment: Assign players to teams based on t-shirt colors using KMeans clustering.

- Ball Assignment: Assign the ball to a player during gameplay.

- Ball Possession Analysis: Measure team ball possession throughout the match.

- Movement Analysis: Track player movements and measure distances in meters using perspective transformation.

## Tech Stack

- Programming Language: Python

- Framework: TensorFlow/PyTorch (for YOLO)

- Libraries: OpenCV, NumPy, scikit-learn, matplotlib, supervision, ultralytics

- Object Detection Model: YOLO (You Only Look Once)

- Clustering Algorithm: KMeans (for team assignment)

- Transformation: Perspective transformation (for real-world measurement)

## Trained YOLO Model Results

  ### Model Summary (Training)

  ![training_res](https://github.com/user-attachments/assets/91d8b3aa-d453-41f5-89b9-4a29849bde57)

  ### Model Summary (Validation)

  ![validation_res](https://github.com/user-attachments/assets/4a4fc3f4-71f7-40f1-ac89-7c9c270a3a3d)

  ### Result

  ![results](https://github.com/user-attachments/assets/1b17aaa4-116a-40d3-9f31-3d540281aa28)

## Final Result (Demo) 

https://github.com/user-attachments/assets/7f43e4f1-dea5-4c83-bf41-fa462d4992b4

## Challenges and Improvements

  

  - Challenges:

    

    * Distinguishing players with similar t-shirt colors.

    * Tracking Players in complex game scenarios.

  - Potential Improvements:

 

    * Training the YOLO with more Data (I only use a small DataSet)

    * Improve goalkepper team asigner

    * Use more advanced clustering techniques for better player-team assignment.