VigilAI

A real-time computer vision system that leverages facial feature analysis and machine learning to detect driver/operator fatigue and prevent accidents through timely alerts

Fatigue Detection System

A real-time computer vision system for detecting driver fatigue using facial landmarks, eye tracking, and drowsiness detection.

Overview

This system monitors a person's face in real-time to detect signs of fatigue by:

• Tracking eye closure patterns
• Calculating Eye Aspect Ratio (EAR)
• Detecting yawning using Mouth Aspect Ratio (MAR)
• Providing real-time audio-visual alerts
• Logging drowsiness events and capturing screenshots

Project Structure

fatigue_detection/
├── data/
│   ├── raw/
│   │   └── CEW Dataset/    # Dataset for validation
│   └── processed/
│       ├── logs/           # Event logs
│       ├── screenshots/    # Drowsiness event captures
│       └── validation/     # Validation results
├── src/
│   └── feature_extraction.py    # Main detection code
├── requirements.txt
└── alert.wav               # Alert sound file

Prerequisites

Hardware Requirements

• Webcam
• Audio output device
• Minimum 4GB RAM
• CPU with SSE4.1 or higher support

Software Requirements

• Python 3.8 or higher
• pip (Python package installer)

Installation

1. Clone the repository:

git clone https://github.com/Aarchishya/fatigue_detection.git
cd fatigue_detection

2. Create and activate a virtual environment:

# Windows
python -m venv venv
venv\Scripts\activate

# Linux/Mac
python -m venv venv
source venv/bin/activate

3. Install required packages:

pip install -r requirements.txt

Required Packages

opencv-python==4.8.0
mediapipe==0.10.0
numpy==1.24.3
scipy==1.11.1
pygame==2.5.0
tqdm==4.65.0

Dataset

For validation, the system uses the CEW (Closed Eyes in the Wild) dataset:

• Download from: https://www2.fpce.uc.pt/~lmv/downloads/databases/CEW_v2/
• Extract to: data/raw/CEW Dataset/
• Structure:

  CEW Dataset/
  ├── Open Eyes/
  └── Closed Eyes/
  

Usage

Real-time Fatigue Detection

python src/feature_extraction.py

Dataset Validation

python src/feature_extraction.py --validate

Controls

• Press 'q' to quit
• Press 'p' to pause/resume
• Press 'm' to toggle metrics display

Features

Real-time Detection

• Face mesh detection using MediaPipe
• Eye closure monitoring
• Yawn detection
• Visual alerts on screen
• Audio alerts for drowsiness

Data Logging

• Timestamps of drowsy events
• EAR and MAR values
• Screenshots of drowsy moments
• Validation metrics

Validation System

• Dataset-based validation
• Accuracy metrics
• Threshold optimization
• Performance analysis

Output Files

Logs

• Location: data/processed/logs/drowsiness_log.csv
• Format: CSV with columns:
  • Timestamp
  • Event Type
  • EAR Value
  • Screenshot Path

Screenshots

• Location: data/processed/screenshots/
• Format: JPEG images
• Naming: drowsy_YYYYMMDD_HHMMSS.jpg

Validation Results

• Location: data/processed/validation/validation_results.txt
• Contains:
  • Accuracy metrics
  • EAR statistics
  • Suggested thresholds

Troubleshooting

1. Camera not detected

   • Check camera connections
   • Verify camera permissions
   • Try different camera index in code

2. Audio alerts not working

   • Check audio device
   • Verify 'alert.wav' exists
   • Check pygame installation

3. High CPU usage

   • Lower camera resolution
   • Increase frame skip
   • Close background applications

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request