ShanmukhaDesu/identix-facial-landmarks

IDENTIX - Facial Landmark Detection & Deepfake Analysis

🚀 Production-Ready Deployment Package

What is IDENTIX?

IDENTIX is an advanced AI-powered web application that performs:

• —Facial Landmark Detection - 11-class segmentation using BiSeNet architecture
• —Deepfake Detection - Multi-factor analysis for video authenticity
• —User Authentication - Secure MongoDB-backed user management
• —Real-time Processing - Image and video analysis with instant results

🎯 Key Features

• —✅ High-accuracy facial landmark segmentation (trained BiSeNet model)
• —✅ Multi-person detection and tracking
• —✅ Deepfake detection with confidence scoring
• —✅ Secure user registration and authentication
• —✅ Beautiful responsive UI with light/dark themes
• —✅ MongoDB Atlas integration for user data
• —✅ Production-ready Flask application
• —✅ Optimized for Render deployment

📋 Quick Start

For deployment instructions, see [DEPLOYMENT.md](DEPLOYMENT.md)

Local Testing

1. 1.Install dependencies:

   pip install -r requirements.txt

1. 1.Set up environment variables:

   cp .env.example .env
   # Edit .env with your MongoDB connection string

1. 1.Run the application:

   python app.py

1. 1.Open browser:

   http://localhost:5000

🏗️ Architecture

Tech Stack

• —Backend: Flask (Python web framework)
• —ML Framework: PyTorch (BiSeNet model)
• —Database: MongoDB Atlas (user authentication)
• —Image Processing: OpenCV, PIL
• —Server: Gunicorn (production)

Model Details

• —Architecture: BiSeNet (Bilateral Segmentation Network)
• —Backbone: ResNet-50
• —Classes: 11 facial landmarks
• —Skin
• —Left Eyebrow
• —Right Eyebrow
• —Left Eye
• —Right Eye
• —Nose
• —Upper Lip
• —Inner Mouth
• —Lower Lip
• —Hair
• —Background

Security Features

• —SHA256 password hashing (upgradable to bcrypt)
• —Email validation with regex
• —Input sanitization (XSS prevention)
• —Unique email constraint in MongoDB
• —Session-based authentication
• —Environment variable configuration

📁 Project Structure

identix-deploy/
├── app.py                  # Main Flask application
├── mongodb_utils.py        # Database connection utilities
├── deepfake_detector.py    # Deepfake analysis module
├── best_model.pth          # Trained BiSeNet model (~95MB)
├── requirements.txt        # Python dependencies
├── render.yaml             # Render deployment config
├── .env.example            # Environment template
├── .gitignore              # Git ignore rules
├── DEPLOYMENT.md           # Detailed deployment guide
├── README.md               # This file
└── templates/              # HTML templates
    ├── index.html          # Landing page
    ├── image_analysis.html # Image upload page
    ├── video_analysis.html # Video upload page
    ├── deepfake_detection.html
    └── ... (other pages)

🔑 Environment Variables

Required variables (see .env.example):

🧪 API Endpoints

Authentication

• —POST /register - Create new user account
• —POST /login - Authenticate user
• —POST /logout - End user session
• —GET /check-auth - Check authentication status

Analysis

• —POST /predict - Analyze image for facial landmarks
• —POST /predict_video - Process video frames
• —POST /detect_deepfake - Analyze video for deepfake

Other

• —GET /health - Health check endpoint
• —GET / - Landing page
• —GET /facial-landmarks - Image analysis UI
• —GET /video-analysis - Video analysis UI

🛡️ Security Considerations

Current Implementation

✅ Password hashing (SHA256) ✅ Input validation ✅ XSS prevention ✅ Secure session management ✅ Environment-based secrets

Recommended Improvements

⚠️ Upgrade to bcrypt/argon2 password hashing ⚠️ Add rate limiting (Flask-Limiter) ⚠️ Add CSRF protection (Flask-WTF) ⚠️ Implement password reset functionality ⚠️ Add email verification

See DEPLOYMENT.md for implementation details.

📊 Performance

Model Performance

• —Input Size: 256×256 pixels
• —Output: 11-class segmentation mask
• —Inference Time: ~100-300ms per image (CPU)

Resource Requirements

• —RAM: ~1GB minimum
• —Storage: ~500MB (model + dependencies)
• —CPU: Single core sufficient for free tier

Render Free Tier

• —750 hours/month
• —Sleeps after 15 minutes inactivity
• —First request after sleep: ~30 seconds

🐛 Known Issues & Limitations

1. 1.Free Tier Sleep: App sleeps on inactivity (Render limitation)
2. 2.Model Size: Large model file (~95MB) affects cold start
3. 3.SHA256 Hashing: Basic password security (upgrade recommended)
4. 4.No Rate Limiting: Vulnerable to brute force (mitigation available)
5. 5.CPU Inference: Slower than GPU (acceptable for free tier)

🔄 Updates & Maintenance

Updating Code

git add .
git commit -m "Update description"
git push origin main
# Render auto-deploys

Monitoring

• —Check Render dashboard for logs
• —Monitor /health endpoint
• —Review MongoDB Atlas metrics

📖 Documentation

• —Deployment Guide: DEPLOYMENT.md
• —Render Setup Summary:
• —Build: pip install --upgrade pip && pip install -r requirements.txt
• —Start: gunicorn app:app -b 0.0.0.0:$PORT -w 2
• —Health: /health
• —Env vars: MONGODB_URI, SECRET_KEY, FLASK_ENV=production
• —Plan: standard recommended for PyTorch model
• —Code Documentation: Inline docstrings in Python files
• —API Reference: See "API Endpoints" section above

🤝 Contributing

This is a capstone project. For modifications:

1. 1.Test locally first
2. 2.Update requirements.txt if adding dependencies
3. 3.Document changes in commit messages
4. 4.Check security implications

📝 License

Educational/Academic Project - Capstone 4-1

👥 Authors

Capstone Team - Facial Landmark Detection Project

🙏 Acknowledgments

• —BiSeNet architecture inspiration
• —MongoDB Atlas free tier
• —Render free hosting
• —PyTorch framework
• —Flask community

Ready to deploy? Follow the comprehensive guide in DEPLOYMENT.md

Questions? Check troubleshooting section in deployment guide.

Need help? Review the security and performance sections above.