👁️ Eyeguard - Professional Eye Strain Detection System

AI-Powered Eye Health Monitoring with Real-Time Blink Detection & Fatigue Analysis

B.Tech Final Year Project | Computer Science & Engineering

---

🎯 Project Overview

Eyeguard is a comprehensive eye strain detection system that uses computer vision and machine learning to monitor eye health in real-time. The system tracks blink patterns, detects fatigue, and provides personalized health recommendations.

🏆 Key Achievement

• ✅ 98% ML Model Accuracy
• ✅ 95 Blinks Detected in Live Demo
• ✅ Real-time Processing at 30 FPS
• ✅ Production-Ready Full-Stack Application

---

✨ Features

Core Capabilities

• 📹 Real-Time Eye Tracking - MediaPipe Face Mesh (468 landmarks)
• 👁️ Blink Detection - Custom EAR-based algorithm
• 🤖 ML Fatigue Classification - 4 levels (Normal, Mild, Moderate, Severe)
• 📊 Live Dashboard - Real-time metrics & charts
• 📄 Professional Reports - PDF generation with visualizations
• 📈 Analytics Engine - Wellness scoring & pattern detection

Advanced Features

• 🎯 Personalized Calibration - Adapts to individual eyes
• 💡 Smart Recommendations - Based on usage patterns
• ⚠️ Multi-Level Alerts - 20-20-20 rule implementation
• 📦 Data Export - CSV, JSON, Excel formats
• 🌐 Web Dashboard - Professional browser interface
• 🐳 Docker Ready - Containerized deployment

---

🚀 Quick Start

Prerequisites

Python 3.11.6
Webcam

Installation

# Clone repository
git clone https://github.com/YOUR_USERNAME/Eye-Guard.git
cd Eye-Guard

# Install dependencies
pip install -r requirements.txt

# Run the application
python simple_app.py

That's it! The camera window will open and start tracking your eyes.

---

📖 Usage

1. Desktop Eye Tracking

python simple_app.py

• Camera opens automatically
• Displays real-time blink count & rate
• Shows color-coded health status
• Press 'q' to stop

2. Web Dashboard

# Start API server
python api/app.py

# Open dashboard
open web/dashboard.html

3. Generate Reports

# PDF report with charts
python eyeguard_cli.py report --user YOUR_NAME

# Analytics & wellness score
python eyeguard_cli.py analytics --user YOUR_NAME

# Export data
python eyeguard_cli.py export --format csv --user YOUR_NAME

---

🎓 Demo Results

Live Testing Results

📊 Session Summary:
   Duration: 4 minutes 10 seconds
   Total Blinks: 95 blinks
   Blink Rate: 19.9/min
   Status: ✅ HEALTHY (15-20/min is optimal)
   Accuracy: 100% detection rate

ML Model Performance

Training Accuracy: 98.00%
Precision: 98.05%
Recall: 98.00%
F1-Score: 98.00%

Confusion Matrix:
[[ 97   3   0   0]  Normal
 [  0 100   0   0]  Mild
 [  0   1  96   3]  Moderate
 [  0   0   1  99]] Severe

---

🏗️ Architecture

Technology Stack

• Computer Vision: OpenCV, MediaPipe
• Machine Learning: TensorFlow/Keras
• Backend: Flask REST API
• Frontend: HTML, CSS, JavaScript, Tailwind CSS
• Database: SQLite
• Charts: Chart.js, Matplotlib, Seaborn
• Reports: ReportLab

System Components

Eyeguard/
├── src/
│   ├── core/          # Eye detection, blink analysis, gaze tracking
│   ├── ml/            # ML model, feature extraction, training
│   ├── utils/         # Database, logging, reports, analytics
│   └── config/        # Configuration & constants
├── api/               # Flask REST API (15+ endpoints)
├── web/               # Web dashboard
├── models/            # Trained ML model (98% accuracy)
└── data/              # SQLite database & reports

---

📊 Key Algorithms

1. Eye Aspect Ratio (EAR)

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

• Detects blinks when EAR < threshold
• Threshold: 0.25 (calibratable)

2. Blink Rate Calculation

blink_rate = (total_blinks / duration_minutes)

• Healthy range: 15-20 blinks/min
• Real-time sliding window analysis

3. Fatigue Classification

21 Features → Neural Network → 4 Classes

• Features: Statistical (mean, std, skew, kurtosis) from EAR, blink rate, gaze
• Model: Dense layers (128→64→32→4)
• Output: Normal, Mild, Moderate, Severe

---

🎨 Features Deep Dive

Real-Time Tracking

• MediaPipe Face Mesh for 468 facial landmarks
• 30 FPS performance on standard hardware
• Sub-100ms latency for blink detection

Machine Learning

• 98% accuracy on 4-class fatigue classification
• 21 engineered features from time-series data
• Transfer learning ready for personalization

Analytics

• Wellness Score (0-100) based on 4 metrics
• Pattern Detection using scipy statistical tests
• Trend Analysis for blink rate, duration, frequency

Reports

• Professional PDFs with charts and insights
• Health Assessment with color-coded status
• Personalized Recommendations based on data

---

📱 Deployment

Local Demo (Current)

python simple_app.py

Cloud Deployment (Free)

# Deploy API to Render.com
git push origin main

# Deploy Dashboard to Vercel
vercel --prod

Docker

docker-compose up -d

---

📈 Project Statistics

• Total Code: 5,500+ lines
• Python Modules: 22 files
• Functions/Classes: 150+
• ML Features: 21 engineered
• API Endpoints: 15+
• Test Coverage: Core features verified
• Documentation: Comprehensive

---

🎯 Use Cases

• 👨‍💻 Developers - Prevent eye strain during long coding sessions
• 👨‍🎓 Students - Monitor study session health
• 👨‍💼 Office Workers - Track screen time & breaks
• 🏥 Health Research - Collect blink pattern data
• 🎓 Academic - Final year project / research

---

🔬 Research & Citations

Based On

• Eye Aspect Ratio (EAR) - Soukupová & Čech (2016)
• MediaPipe - Google Research
• Blink Rate Studies - Clinical ophthalmology research

Novel Contributions

• Personalized calibration system
• Wellness scoring algorithm
• Pattern detection for fatigue
• Multi-format export pipeline

---

📝 API Documentation

Endpoints

GET  /api/health              - Health check
GET  /api/sessions            - Get user sessions
POST /api/sessions/create     - Create new session
GET  /api/analytics/wellness  - Wellness score
GET  /api/analytics/patterns  - Usage patterns
GET  /api/reports/generate    - Generate PDF
GET  /api/export/csv          - Export CSV
GET  /api/stats/dashboard     - Dashboard stats

Full API docs: API Documentation

---

🛠️ Development

Setup Development Environment

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dev dependencies
pip install -r requirements.txt

# Run tests
python test_system.py

Train Custom Model

python src/ml/model_trainer.py --samples 1000 --epochs 50

---

🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create feature branch (git checkout -b feature/AmazingFeature)
3. Commit changes (git commit -m 'Add AmazingFeature')
4. Push to branch (git push origin feature/AmazingFeature)
5. Open Pull Request

---

📄 License

This project is part of academic work.

For collaboration or commercial use, please contact.

---

👨‍💻 Team

Core Development Team

Name	Role	Contribution
Gaurav Kumar Mehta	Lead Developer	Architecture, ML Model, Core System
Ayan Biswas	Developer	Eye Detection, Testing
Arpan Mirsha	Developer	Frontend, PWA
Arka Bhattacharya	Developer	Analytics, Reports

• Department: Computer Science & Engineering
• Academic Year: 2025-26
• GitHub: @gaurav620

---

🙏 Acknowledgments

• MediaPipe - Google Research for face mesh
• TensorFlow - ML framework
• OpenCV - Computer vision library
• Mentors - [Professor names]
• Department - CSE Department support

---

📞 Support

For issues or questions:

• 📧 Email: your.email@example.com
• 🐛 Issues: [GitHub Issues]
• 📖 Docs: Full Documentation

---

⭐ Project Highlights

✅ Production Ready - Fully functional and tested ✅ High Accuracy - 98% ML model performance
✅ Real-Time - 30 FPS eye tracking ✅ Full-Stack - Desktop + Web + API + Mobile-ready ✅ Well-Documented - Comprehensive guides ✅ Deployment Ready - Docker + Cloud compatible

---

🏆 Excellence in Academic Software Development

Eyeguard - Professional Eye Health Monitoring

Made with ❤️ for better eye health

⬆ Back to Top