🎯 AI-Powered Football Match Analyzer
A real-time system that uses deep learning and computer vision to analyze football matches, track players and the ball, and provide instant insights to support coaches in decision-making.
📂 Project Structure
players_detect_project/
│
├── camera_movement_estimator/ # Estimate camera movement
├── development_and_analysis/ # Data cleaning and visual analysis
├── player_ball_assigner/ # Assign ball possession to player
├── speed_and_distance_estimator/ # Calculate player speed and distance
├── stubs/ # Preprocessed files (Pickle)
├── team_assigner/ # Detect team for each player
├── trackers/ # Object tracking logic
├── view_transformer/ # Convert camera view to top-down
├── best.pt # Trained YOLOv8 model weights
├── main.py # Main entry to run the full pipeline
└── yolo_inference.py # Inference script using YOLOv8
🧠 How It Works
- Video Input: The system accepts football match videos.
- YOLOv8 Inference: Detects players, referees, ball, and goalkeeper.
- Tracking & Assignment:
- Tracks players and ball movement over time.
- Assigns ball possession to the nearest player.
- Assigns team labels and colors.
- Analytics Modules:
- Calculates player speed, distance, and performance drops.
- Detects camera movement to adjust accuracy.
- Sends real-time alerts to the coach when tactics shift or performance declines.
✅ Key Features
- Real-time AI-based decision support for coaches.
- Precision tracking of players, ball, and game flow.
- Insights like ball possession %, sprint speed, tactical shifts.
- Modular design, easily extendable for future improvements.
🧪 Model Performance
| Class | Accuracy |
|---|---|
| Player | 99.4% |
| Goalkeeper | 94.8% |
| Referee | 97.7% |
| Ball | 59.9% ⚠️ |
Tested on 50 real match clips. Ball detection will be improved with more data.
🚧 Future Plans
- Improve ball detection accuracy with more diverse data.
- Add user interface for live visualization.
- API integration for live camera input.
- Full match testing in real-world environments.
🤝 Contributors & Support
Built as part of the SDAIA AI League Hackathon, with the goal of empowering sports strategy using AI.