⚽ Football Analysis with Computer Vision

A complete end-to-end football (soccer) analysis pipeline built using YOLO-based tracking, camera motion compensation, homography transformations, and player speed/distance estimation.

This project takes raw broadcast video as input and outputs an annotated video with:

• Player & ball tracking
• Team assignment
• Ball possession
• Player speed (km/h) & distance (meters)
• Camera movement visualization

---

🚀 Features

✅ Object Tracking – Players, ball, and referees tracked using YOLO + SORT/ByteTrack.
✅ Camera Motion Compensation – Adjusts for panning/zooming cameras to keep positions consistent.
✅ View Transformation (Homography) – Converts pixel coordinates to real-world field coordinates (105 × 68 m).
✅ Player Speed & Distance Estimation – Calculates realistic sprint speeds and distances covered.
✅ Team Assignment – Classifies players into two teams based on jersey color.

---

📂 Project Structure

.
├── config.py                # Configuration (paths, model, constants)
├── main.py                  # Main entry point
├── trackers/                # Object tracking (players, ball, referees)
├── camera_movement_estimator.py  
├── view_transformer.py      # Homography transformation (pixel → meters)
├── speed_distance_estimator.py  
├── team_assigner.py         
├── player_ball_assigner.py  
├── utils.py                 # Helper functions (video I/O, distance, etc.)
├── images/                  # Project images & figures
│   └── img.png
└── README.md

---

⚙️ Installation

1. Clone the repository:

git clone https://github.com/AhmedQassemDev2004/computer-vision-football-analysis/
cd football_analysis

2. Install dependencies:

pip install -r requirements.txt

3. Download 'best.pt' model weights and place them in models/ (update config.py with the correct path).

---

▶️ Usage

Run the pipeline:

python main.py

Input video: input_videos/match.mp4
Output video: output_videos/match_annotated.mp4

---

📊 Output Annotations

• Player IDs, team colors, and possession info
• Player speed (km/h) and distance covered (m)
• Ball position and assignment to players
• Camera motion overlay

---

🛠️ Future Improvements

• Improve Players distance and speed detection
• Better Goalkeeper to team assigment ( Temporarily hard coded untill fix )
• Improve homography calibration (automatic line detection)
• Add pass & event detection (e.g., shots, tackles)
• Integrate real match stats export (CSV/JSON)