Sleep Detector for Drivers

A real-time computer vision application that detects driver drowsiness using facial landmarks and eye aspect ratio analysis. When drowsiness is detected, the system triggers an audio alarm to alert the driver.

Features

• Real-time eye detection using webcam
• Drowsiness detection based on eye aspect ratio (EAR)
• Audio alarm when driver shows signs of drowsiness
• Visual indicators (color-coded eye rectangles)
• Adjustable sensitivity thresholds

Prerequisites

Before running this project, ensure you have:

• Python 3.7+ installed on your system
• A working webcam
• An audio file named alarm.mp3 in the project directory

Installation

1. Clone or Download the Project

cd Sleep-Detector

2. Install Required Dependencies

Run the following command to install all required Python packages:

pip install opencv-python dlib scipy imutils pygame

Package Descriptions:

• opencv-python - Computer vision library for video capture and image processing
• dlib - Machine learning library for facial landmark detection
• scipy - Scientific computing library (for distance calculations)
• imutils - OpenCV convenience functions
• pygame - For playing audio alarms

3. Download the Face Landmarks Model

Download the shape_predictor_68_face_landmarks.dat file from:

• iBUG 300-W Dataset

Installation steps:

1. Download and extract the .bz2 file
2. Place the shape_predictor_68_face_landmarks.dat file in your project directory
3. Update the path in main.py if needed (currently set to: C:\Users\yadav\OneDrive\Desktop\day 17\shape_predictor_68_face_landmarks.dat)

4. Add Your Alarm Audio File

Add an audio file named alarm.mp3 to the project directory. This will play when drowsiness is detected.

Configuration

You can adjust the following parameters in main.py:

ALARM_FILE = "alarm.mp3"                      # Path to your alarm audio file
EYE_ASPECT_RATIO_THRESHOLD = 0.25             # EAR threshold (lower = more sensitive)
EYE_ASPECT_RATIO_CONSEC_FRAMES = 20           # Frames threshold before triggering alarm

Adjusting Sensitivity:

• Lower EYE_ASPECT_RATIO_THRESHOLD = More sensitive (detects drowsiness faster)
• Higher EYE_ASPECT_RATIO_THRESHOLD = Less sensitive (requires more obvious eye closure)
• EYE_ASPECT_RATIO_CONSEC_FRAMES = Number of consecutive frames with closed eyes before alarm triggers

Running the Application

1. Open a terminal/command prompt in the project directory
2. Run the application:

python main.py

3. The webcam feed will open in a new window showing:

   • Green rectangles around eyes = Eyes are open (normal)
   • Red rectangles around eyes = Drowsiness detected (alarm will sound)

4. To exit the application, press Q on your keyboard

How It Works

1. Face Detection - Uses dlib's frontal face detector to locate faces in the video stream
2. Facial Landmarks - Detects 68 facial landmarks including eye coordinates
3. Eye Aspect Ratio (EAR) - Calculates the ratio of vertical to horizontal eye distances
4. Drowsiness Detection - If EAR falls below the threshold for consecutive frames, drowsiness is detected
5. Alarm Trigger - An audio alarm plays and a visual alert appears on screen

Troubleshooting

Common Issues

Issue: "shape_predictor_68_face_landmarks.dat not found"

• Solution: Download the file and update the path in main.py line 26

Issue: "No module named 'dlib'"

• Solution: Install dlib using pip install dlib (may take a few minutes to compile)

Issue: Alarm doesn't play

• Solution: Ensure alarm.mp3 exists in the project directory and pygame is properly installed

Issue: Webcam not detected

• Solution: Check if another application is using your webcam or try changing the camera index from 0 to 1 in main.py line 34

Issue: Low FPS or laggy detection

• Solution: The video is resized to 960x720. You can reduce this for better performance or increase for better quality

Project Structure

Sleep-Detector/
├── main.py                                    # Main application file
├── README.md                                  # This file
├── alarm.mp3                                  # Your alarm audio file
└── shape_predictor_68_face_landmarks.dat      # Face landmarks model

License

This project is open source and available under the MIT License.

Author

Created as a driver safety application to prevent accidents caused by driver drowsiness.