🎥 AI CCTV Face Recognition System

A production-ready, cloud-deployable face recognition system using InsightFace's SCRFD detector and ArcFace recognition model. Designed for CCTV monitoring, video analysis, and real-time face matching.

✨ Features

• 🎯 High Accuracy: SCRFD detector + ArcFace recognition (state-of-the-art)
• 🌐 Web Interface: Modern, intuitive UI for uploading photos/videos and viewing results
• ☁️ Cloud Ready: Docker containerized for easy deployment to AWS, GCP, Azure, Railway, etc.
• 📹 Multiple Input Sources: Images, video files, RTSP CCTV feeds
• 💾 Flexible Database: SQLite (default), PostgreSQL, Redis support
• ⚡ Real-Time Processing: Fast cosine similarity matching (100k+ embeddings in <5ms)
• 🔔 Event System: Alerts, logs, tracking, webhooks
• 📊 Dashboard: Real-time monitoring and statistics

🚀 Quick Start

Option 1: Web Interface (Recommended)

1. Start the web application:

python web_app.py

2. Open browser: http://localhost:5000

3. Upload blacklist photos in the "Blacklist Management" tab

4. Process videos in the "Process Video" tab (upload or provide file path)

Option 2: Command Line

1. Build watchlist from photos:

python main.py build-watchlist --photos-dir ./photos --output-db watchlist.db

2. Process video:

python main.py process-video --video ./demo.mp4 --watchlist watchlist.db --output results.json

3. Process RTSP stream:

python main.py process-rtsp --rtsp-url rtsp://camera-url --watchlist watchlist.db

☁️ Cloud Deployment

Docker (Recommended)

# Build and run
docker-compose up -d

# Or manually
docker build -t ai-cctv-face-recognition .
docker run -p 5000:5000 ai-cctv-face-recognition

Supported Platforms

• ✅ AWS (EC2, ECS, Elastic Beanstalk)
• ✅ Google Cloud (Cloud Run, Compute Engine)
• ✅ Azure (Container Instances, App Service)
• ✅ DigitalOcean (App Platform, Droplets)
• ✅ Railway
• ✅ Render
• ✅ Any Docker-compatible platform

See DEPLOYMENT.md for detailed cloud deployment instructions.

Architecture

• face_detector.py: SCRFD face detection
• face_recognizer.py: ArcFace embedding generation
• database.py: Database abstraction layer
• matcher.py: Real-time face matching
• event_handler.py: Event system (alerts, logs, webhooks)
• input_handlers.py: Image, video, RTSP input processing
• dashboard.py: Web UI dashboard

Configuration

Create a .env file:

SIMILARITY_THRESHOLD=0.35
DATABASE_TYPE=sqlite
DATABASE_URL=watchlist.db
REDIS_HOST=localhost
REDIS_PORT=6379
WEBHOOK_URL=https://your-webhook-url.com