Football Analysis via ML

This project aims to perform advanced football video analysis using cutting-edge machine learning and computer vision techniques. The main objectives include detecting and tracking players, referees, and footballs in video footage, assigning players to teams based on their t-shirt colors, and analyzing player movements and performance metrics.

---

Key Features

1. Object Detection:

   • Detect and track players, referees, and footballs using YOLO, one of the leading AI object detection models.
   • Train YOLO for improved detection performance on football-related objects.

2. Team Assignment:

   • Assign players to teams by analyzing t-shirt colors using KMeans clustering for pixel segmentation.
   • Calculate team ball possession percentages during matches.

3. Camera Motion Estimation:

   • Use optical flow to measure camera movement between frames, enabling accurate player movement analysis.

4. Perspective Transformation:

   • Apply perspective transformation to account for depth and perspective, allowing movement measurements in meters rather than pixels.

5. Performance Metrics:

   • Calculate individual player speeds and distances covered during the game.

---

Modules Used

This project employs the following modules:

• YOLO: AI object detection model.
• KMeans: Pixel segmentation and clustering to determine team t-shirt colors.
• Optical Flow: Analyze camera movement.
• Perspective Transformation: Measure scene depth and perspective.
• Custom Utilities: Compute player speed and distance.

Requirements

Ensure you have the following dependencies installed:

• Python 3.x
• ultralytics
• supervision
• OpenCV
• NumPy
• Matplotlib
• Pandas