🔐 Real-Time Biometric Authentication System

🚀 Production-grade facial recognition system for real-time identity verification using computer vision and machine learning

---

🌟 Key Highlights

• 🎯 High-Accuracy Face Recognition using HOG + SVM pipeline
• ⚡ Real-Time Inference optimized for low-latency video streams
• 🐳 Dockerized Deployment for reproducible environments
• 🧱 Modular Architecture (Detection → Feature Extraction → Classification → Serving)
• 🎨 Interactive Dashboard with modern UI
• 📈 Designed for scalability and extensibility

---

🧠 System Architecture

Video Stream (Webcam)
        ↓
Face Detection (Haar Cascade / DNN)
        ↓
Preprocessing (Grayscale + Resize 64x64)
        ↓
Feature Extraction (HOG)
        ↓
Classification (SVM - RBF Kernel)
        ↓
Identity + Confidence Score

---

⚙️ Tech Stack

Layer	Technology
Backend	Flask (Python 3.9+)
Computer Vision	OpenCV, Scikit-Image
ML Model	Scikit-Learn (SVM, HOG)
Model Persistence	Joblib
Deployment	Docker, Docker Compose

---

🚀 Quick Start

🔹 Local Setup

python -m venv venv

# Windows
venv\Scripts\activate

# macOS/Linux
source venv/bin/activate

pip install -r requirements.txt

python -m app.main

👉 Open: http://localhost:5000

---

🐳 Docker Setup

docker-compose up --build

⚠️ Webcam support in Docker:

• Linux → supported via --device
• Windows/Mac → recommended to run locally

---

🧪 Model Training

Dataset Structure

dataset/
├── Alice/
├── Bob/

• Add 10–20 images per person
• Use varied angles and lighting

Train Model

python model/train_model.py

➡️ Outputs: model/svm_model.pkl

---

👤 User Enrollment (Real-Time)

python scripts/enroll_user.py

Steps:

1. Enter user name
2. Capture 10–20 images
3. Auto-trains model
4. Restart app

---

📊 Performance

Metric	Value
Latency	Real-time (~30 FPS depending on hardware)
Accuracy	High (depends on dataset quality)
Model Type	HOG + SVM (RBF Kernel)
Scalability	Medium (can upgrade to deep learning)

---

🧠 Design Decisions

• HOG + SVM chosen for:

  • Fast inference
  • Low compute requirement
  • Works well on small datasets

• Flask Backend:

  • Lightweight API serving
  • Easy integration with UI

• Docker:

  • Consistent deployment
  • Environment isolation

---

🔮 Future Improvements

• Replace SVM with FaceNet / DeepFace embeddings
• Add FAISS / Vector DB for scalable identity matching
• Implement liveness detection
• Add JWT-based auth system
• GPU acceleration for inference

---

📂 Project Structure

biometric-auth-system/
├── app/
├── model/
├── scripts/
├── dataset/
├── requirements.txt
└── docker-compose.yml

---

👤 Author

Raj Kalash Tiwari GitHub: https://github.com/rjkalash

---

📈 Status

✅ Functional real-time biometric authentication system ⚡ Ready for extension into production-grade identity systems

---

⭐ Star this repo if you found it useful!