Driver Drowsiness Detection System

A real-time system for detecting driver drowsiness using computer vision and facial landmarks detection. This system monitors a driver's eyes and alerts them when signs of drowsiness are detected.

Features

• Real-time face detection using Haar Cascades and dlib
• Eye state monitoring using Eye Aspect Ratio (EAR)
• Facial landmarks detection for precise eye tracking
• Audio alerts when drowsiness is detected
• Multiple detection methods (Haar Cascade and dlib) for comparison
• Face direction monitoring for distraction detection
• Yawn detection for additional drowsiness indicators

Project Structure

driver-drowsiness-detection/
├── data/
│   ├── alert_sounds/         # Alert sound files
│   │   ├── alert-sound.mp3
│   │   ├── focus.mp3
│   │   └── take_a_break.mp3
│   ├── database/            # Shape predictor data file
│   │   └── shape_predictor_68_face_landmarks.dat
│   ├── images/             # Output directory for eye images
│   └── xml/                # Haar cascade classifier files
│       ├── haarcascade_eye.xml
│       ├── haarcascade_frontalface_alt.xml
│       ├── haarcascade_frontalface_alt2.xml
│       ├── haarcascade_frontalface_alt_tree.xml
│       └── haarcascade_frontalface_default.xml
└── src/                    # Source code
    ├── final_integration.py  # Main implementation with all features
    ├── main.py              # Basic implementation using Haar cascades
    ├── main_dlib.py         # Implementation using dlib
    └── train.py             # Training utility for eye state detection

Dependencies

• Python 3.x
• OpenCV (cv2)
• dlib
• numpy
• imutils
• matplotlib
• python-vlc

Installation

1. Clone the repository:

git clone https://github.com/LohiyaH/DrowsinessGuard.git
cd driver-drowsiness-detection

2. Install the required dependencies:

pip install opencv-python dlib numpy imutils matplotlib python-vlc

3. Download the shape predictor file:

• Download shape_predictor_68_face_landmarks.dat from dlib's official website
• Place it in the data/database/ directory

Usage

The project provides three different implementations:

1. Basic Implementation (main.py)

python src/main.py

• Uses Haar cascades for face and eye detection
• Saves detected eye images to data/images/

2. Dlib Implementation (main_dlib.py)

python src/main_dlib.py

• Uses dlib for facial landmark detection
• More accurate than Haar cascade method
• Includes basic drowsiness detection

3. Full Featured Implementation (final_integration.py)

python src/final_integration.py

• Combines all features including:
  • Face direction monitoring
  • Yawn detection
  • Advanced drowsiness detection
  • Multiple alert thresholds
  • Audio alerts

How It Works

1. Face Detection: The system first detects the driver's face using either Haar cascades or dlib's face detector.

2. Facial Landmarks: Using dlib's facial landmark predictor, 68 points on the face are detected, including 6 points for each eye.

3. Eye Aspect Ratio (EAR): The system calculates the eye aspect ratio to determine if the eyes are closed:

   EAR = (||p2-p6|| + ||p3-p5||) / (2||p1-p4||)
   

   where p1, ..., p6 are the 2D landmark points around the eye.

4. Drowsiness Detection: If the EAR falls below a threshold for a certain number of consecutive frames, the system triggers an alert.

5. Additional Features:

   • Face direction monitoring to detect distraction
   • Yawn detection using mouth aspect ratio
   • Multiple alert levels based on drowsiness severity

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is under MIT License.

Acknowledgments

• The shape predictor model is provided by dlib
• Haar cascade classifiers are from OpenCV's repository