Face Recognition System

A deep learning-based face recognition system with web interface for photo management and label control.

Presentation is available at

https://docs.google.com/presentation/d/11goJALVAzFKwUOgDDz5O3_1OvpuqU9Wsu3-GoWeNxQs/edit?usp=sharing

Features

• Face detection using MTCNN
• Face recognition using FaceNet embeddings
• KNN classification for face matching
• Real-time processing
• Label management system
• Web-based user interface

Technical Stack

• Backend: Flask, PyTorch, FaceNet
• Frontend: React, Vite
• Database: SQLite
• Containerization: Podman

Prerequisites

• Python 3.9+
• Node.js 16+
• Podman
• Pre-trained face recognition model (PKL file)

Project Structure

face-recognition-app/
├── backend/
│   ├── embeddings/
│   │   └── face_recognition_model.pkl
│   ├── instance/
│   ├── static/
│   │   └── uploaded_photos/
│   ├── app.py
│   ├── requirements.txt
│   └── Containerfile
├── frontend/
│   ├── src/
│   ├── package.json
│   ├── nginx.conf
│   └── Containerfile
└── README.md

Model Details

The system uses:

• MTCNN for face detection
• FaceNet (InceptionResNetV1) for feature extraction
• KNN classifier for face recognition
• Pre-trained on VGGFace2 dataset

Deployment Instructions

Local Development

1. Clone repository:

git clone <repository-url>
cd face-recognition-app

2. Backend setup:

cd backend
python -m venv facenv
source facenv/bin/activate  # or `facenv\Scripts\activate` on Windows
pip install -r requirements.txt
python app.py

3. Frontend setup:

cd frontend
npm install
npm run dev

Podman Deployment

1. Build Images:

# Backend
cd backend
podman build -t face-recognition-backend .

# Frontend
cd frontend
podman build -t face-recognition-frontend .

2. Create Pod and Run Containers:

# Create pod
podman pod create --name face-recognition-app -p 5000:5000 -p 3000:80

# Run backend
podman run -d --pod face-recognition-app \
    --name backend \
    -v ./backend/static/uploaded_photos:/app/static/uploaded_photos:Z \
    -v ./backend/instance:/app/instance:Z \
    -v ./backend/embeddings:/app/embeddings:Z \
    face-recognition-backend

# Run frontend
podman run -d --pod face-recognition-app \
    --name frontend \
    face-recognition-frontend

3. Check Status:

podman pod ps
podman logs backend
podman logs frontend

Remote Deployment

1. Transfer files:

rsync -avz \
    --exclude 'node_modules' \
    --exclude '.git' \
    --exclude '__pycache__' \
    --exclude 'dist' \
    --exclude 'facenv' \
    --exclude '*.pyc' \
    ./ user@remote-ip:~/face-recognition-app/

2. SSH and deploy:

ssh user@remote-ip
cd face-recognition-app
# Follow Podman deployment steps above

API Endpoints

• POST /upload: Upload photos
• GET /photos: List all photos
• POST /detect: Detect faces
• PUT /label: Update labels

Environment Variables

Backend:

FLASK_ENV=development
FLASK_APP=app.py

Frontend:

VITE_API_URL=http://localhost:5000

Performance

Classification Report:

              precision  recall  f1-score  support
fidan           0.78     0.80     0.79      59
leyla           0.72     0.90     0.80      58
others          0.58     0.26     0.36      27

Troubleshooting

Common issues:

1. Port conflicts: Change ports in pod creation
2. Permission issues: Add :Z to volume mounts
3. Model not found: Check embeddings directory
4. Container fails: Check logs with podman logs

Contributing

MIT License

License

Elnur Asgarli