PunimTag Web

Modern Photo Management and Facial Recognition System

A fast, simple, and modern web application for organizing and tagging photos using state-of-the-art DeepFace AI with ArcFace recognition model.

---

🎯 Features

• 🌐 Web-Based: Modern React frontend with FastAPI backend
• 🔥 DeepFace AI: State-of-the-art face detection with RetinaFace and ArcFace models
• 🎯 Superior Accuracy: 512-dimensional embeddings (4x more detailed than face_recognition)
• ⚙️ Multiple Detectors: Choose from RetinaFace, MTCNN, OpenCV, or SSD detectors
• 🎨 Flexible Models: Select ArcFace, Facenet, Facenet512, or VGG-Face recognition models
• 👤 Person Identification: Identify and tag people across your photo collection
• 🤖 Smart Auto-Matching: Intelligent face matching with quality scoring and cosine similarity
• 🔍 Advanced Search: Search by people, dates, tags, and folders
• 🏷️ Tag Management: Organize photos with hierarchical tags
• ⚡ Batch Processing: Process thousands of photos efficiently
• 🔒 Privacy-First: All data stored locally, no cloud dependencies

---

🚀 Quick Start

Prerequisites

• Python 3.12 or higher
• Node.js 18+ and npm
• Virtual environment (recommended)

Installation

# Clone the repository
git clone <repository-url>
cd punimtag

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

# Install Python dependencies
pip install -r requirements.txt

# Install frontend dependencies
cd frontend
npm install
cd ..

Database Setup

Automatic Initialization: The database and all tables are automatically created on first startup. No manual migration is needed!

The web application will:

• Create the database file at data/punimtag.db (SQLite default) if it doesn't exist
• Create all required tables with the correct schema on startup
• Match the desktop version schema exactly for compatibility

Manual Setup (Optional): If you need to reset the database or create it manually:

source venv/bin/activate
export PYTHONPATH=/home/ladmin/Code/punimtag
# Recreate all tables from models
python scripts/recreate_tables_web.py

PostgreSQL (Production): Set the DATABASE_URL environment variable:

export DATABASE_URL=postgresql+psycopg2://user:password@host:port/database

Database Schema: The web version uses the exact same schema as the desktop version for full compatibility:

• photos - Photo metadata (path, filename, date_taken, processed)
• people - Person records (first_name, last_name, middle_name, maiden_name, date_of_birth)
• faces - Face detections (encoding, location, quality_score, face_confidence, exif_orientation)
• person_encodings - Person face encodings for matching
• tags - Tag definitions
• phototaglinkage - Photo-tag relationships (with linkage_type)

Running the Application

Prerequisites:

• Redis must be installed and running (for background jobs)

  Install Redis (if not installed):

  # On Ubuntu/Debian:
  sudo apt update && sudo apt install -y redis-server
  sudo systemctl start redis-server
  sudo systemctl enable redis-server  # Auto-start on boot
  
  # On macOS with Homebrew:
  brew install redis
  brew services start redis
  
  # Verify Redis is running:
  redis-cli ping  # Should respond with "PONG"
  

  Start Redis (if installed but not running):

  # On Linux:
  sudo systemctl start redis-server
  
  # Or run directly:
  redis-server
  

Terminal 2 - Backend API (automatically starts RQ worker):

cd /home/ladmin/Code/punimtag
source venv/bin/activate
export PYTHONPATH=/home/ladmin/Code/punimtag
uvicorn src.web.app:app --host 127.0.0.1 --port 8000

You should see:

✅ RQ worker started in background subprocess (PID: ...)
INFO:     Started server process
INFO:     Uvicorn running on http://127.0.0.1:8000

Note: The RQ worker automatically starts in a background subprocess when the API starts. You'll see a confirmation message with the worker PID. If Redis isn't running, you'll see a warning message.

Terminal 3 - Frontend:

cd /home/ladmin/Code/punimtag/frontend
npm run dev

Then open your browser to http://localhost:3000

Default Login:

• Username: admin
• Password: admin

Note:

• The database and tables are automatically created on first startup - no manual setup needed!
• The RQ worker starts automatically in a background subprocess when the API server starts
• Make sure Redis is running first, or the worker won't start
• Worker names are unique to avoid conflicts when restarting
• Photo uploads are stored in data/uploads (configurable via PHOTO_STORAGE_DIR env var)
• DeepFace models download automatically on first use (can take 5-10 minutes)
• If port 8000 is in use, kill the process: lsof -i :8000 then kill <PID> or pkill -f "uvicorn.*app"

---

📖 Documentation

• Architecture: System design and technical details
• Web Migration Plan: Detailed migration roadmap
• Phase 1 Status: Phase 1 implementation status
• Phase 1 Checklist: Complete Phase 1 checklist

Phase 2 Features:

• Photo import via folder scan or file upload
• Background processing with progress tracking
• Real-time job status updates (SSE)
• Duplicate detection by checksum
• EXIF metadata extraction
• DeepFace face detection and recognition pipeline
• Configurable detectors (RetinaFace, MTCNN, OpenCV, SSD)
• Configurable models (ArcFace, Facenet, Facenet512, VGG-Face)
• Process tab UI for face processing
• Job cancellation support

---

🏗️ Project Structure

punimtag/
├── src/                    # Source code
│   ├── web/               # Web backend
│   │   ├── api/           # API routers
│   │   ├── db/            # Database models and session
│   │   ├── schemas/       # Pydantic models
│   │   └── services/      # Business logic services
│   └── core/              # Legacy desktop business logic
├── frontend/               # React frontend
│   ├── src/
│   │   ├── api/           # API client
│   │   ├── components/    # React components
│   │   ├── context/       # React contexts (Auth)
│   │   ├── hooks/         # Custom hooks
│   │   └── pages/         # Page components
│   └── package.json
├── tests/                  # Test suite
├── docs/                   # Documentation
├── data/                   # Application data (database, images)
├── alembic/                # Database migrations
└── deploy/                 # Docker deployment configs

---

📊 Current Status

Phase 1: Foundations ✅ COMPLETE

Backend:

• ✅ FastAPI application with CORS middleware
• ✅ Health, version, and metrics endpoints
• ✅ JWT authentication (login, refresh, user info)
• ✅ Job management endpoints (RQ/Redis integration)
• ✅ SQLAlchemy models for all entities
• ✅ Alembic migrations configured and applied
• ✅ Database initialized (SQLite default, PostgreSQL supported)
• ✅ RQ worker auto-start (starts automatically with API server)

Frontend:

• ✅ React + Vite + TypeScript setup
• ✅ Tailwind CSS configured
• ✅ Authentication flow with login page
• ✅ Protected routes with auth context
• ✅ Navigation layout (left sidebar + top bar)
• ✅ All page routes (Dashboard, Scan, Process, Search, Identify, Auto-Match, Tags, Settings)

Database:

• ✅ All tables created automatically on startup: photos, faces, people, person_encodings, tags, phototaglinkage
• ✅ Schema matches desktop version exactly for full compatibility
• ✅ Indices configured for performance
• ✅ SQLite database at data/punimtag.db (auto-created if missing)

Phase 2: Image Ingestion & Processing ✅ COMPLETE

Backend:

• ✅ Photo import service with checksum computation
• ✅ EXIF date extraction and image metadata
• ✅ Folder scanning with recursive option
• ✅ File upload support
• ✅ Background job processing with RQ
• ✅ Real-time job progress via SSE (Server-Sent Events)
• ✅ Duplicate detection (by path and checksum)
• ✅ Photo storage configuration (PHOTO_STORAGE_DIR)
• ✅ DeepFace pipeline integration
• ✅ Face detection (RetinaFace, MTCNN, OpenCV, SSD)
• ✅ Face embeddings computation (ArcFace, Facenet, Facenet512, VGG-Face)
• ✅ Face processing service with configurable detectors/models
• ✅ EXIF orientation handling
• ✅ Face quality scoring and validation
• ✅ Batch processing with progress tracking
• ✅ Job cancellation support

Frontend:

• ✅ Scan tab UI with folder selection
• ✅ Drag-and-drop file upload
• ✅ Recursive scan toggle
• ✅ Real-time job progress with progress bar
• ✅ Job status monitoring (SSE integration)
• ✅ Results display (added/existing counts)
• ✅ Error handling and user feedback
• ✅ Process tab UI with configuration controls
• ✅ Detector/model selection dropdowns
• ✅ Batch size configuration
• ✅ Start/Stop processing controls
• ✅ Processing progress display with photo count
• ✅ Results summary (faces detected, faces stored)
• ✅ Job cancellation support

Worker:

• ✅ RQ worker auto-starts with API server
• ✅ Unique worker names to avoid conflicts
• ✅ Graceful shutdown handling
• ✅ String-based function paths for reliable serialization

Next: Phase 3 - Identify Workflow & Auto-Match

• Identify workflow UI
• Auto-match engine with similarity thresholds
• Unidentified faces management
• Person creation and linking
• Batch identification support

---

🔧 Configuration

Database

SQLite (Default for Development):

# Default location: data/punimtag.db
# No configuration needed

PostgreSQL (Production):

export DATABASE_URL=postgresql+psycopg2://user:password@host:port/database

Environment Variables

# Database (optional, defaults to SQLite)
DATABASE_URL=sqlite:///data/punimtag.db

# JWT Secrets (change in production!)
SECRET_KEY=your-secret-key-here

# Single-user credentials (change in production!)
ADMIN_USERNAME=admin
ADMIN_PASSWORD=admin

# Photo storage directory (default: data/uploads)
PHOTO_STORAGE_DIR=data/uploads

---

🧪 Testing

# Backend tests (to be implemented)
cd /home/ladmin/Code/punimtag
source venv/bin/activate
export PYTHONPATH=/home/ladmin/Code/punimtag
pytest tests/

# Frontend tests (to be implemented)
cd frontend
npm test

---

🗺️ Roadmap

✅ Phase 1: Foundations (Complete)

• FastAPI backend scaffold
• React frontend scaffold
• Authentication system
• Database setup
• Basic API endpoints

✅ Phase 2: Image Ingestion & Processing (Complete)

• ✅ Photo import (folder scan and file upload)
• ✅ Background job processing with RQ
• ✅ Real-time progress tracking via SSE
• ✅ Scan tab UI implementation
• ✅ Duplicate detection and metadata extraction
• ✅ DeepFace face detection and processing pipeline
• ✅ Process tab UI with configuration controls
• ✅ Configurable detectors and models
• ✅ Face processing with progress tracking
• ✅ Job cancellation support

🔄 Phase 3: Identify Workflow & Auto-Match (In Progress)

• Identify workflow UI
• Auto-match engine with similarity thresholds
• Unidentified faces management
• Person creation and linking

📋 Phase 4: Search & Tags

• Search endpoints with filters
• Tag management UI
• Virtualized photo grid
• Advanced filtering

🎨 Phase 5: Polish & Release

• Performance optimization
• Accessibility improvements
• Production deployment
• Documentation

---

🏗️ Architecture

Backend:

• Framework: FastAPI (Python 3.12+)
• Database: SQLite (dev), PostgreSQL (production)
• ORM: SQLAlchemy 2.0
• Migrations: Alembic
• Jobs: Redis + RQ
• Auth: JWT (python-jose)

Frontend:

• Framework: React 18 + TypeScript
• Build Tool: Vite
• Styling: Tailwind CSS
• State: React Query + Context API
• Routing: React Router

Deployment:

• Docker Compose for local development
• Containerized services for production

---

📦 Dependencies

Backend:

• fastapi==0.115.0
• uvicorn[standard]==0.30.6
• pydantic==2.9.1
• SQLAlchemy==2.0.36
• alembic==1.13.2
• python-jose[cryptography]==3.3.0
• redis==5.0.8
• rq==1.16.2
• deepface>=0.0.79
• tensorflow>=2.13.0

Frontend:

• react==18.2.0
• react-router-dom==6.20.0
• @tanstack/react-query==5.8.4
• axios==1.6.2
• tailwindcss==3.3.5

---

🔒 Security

• JWT-based authentication with refresh tokens
• Password hashing (to be implemented in production)
• CORS configured for development (restrict in production)
• SQL injection prevention via SQLAlchemy ORM
• Input validation via Pydantic schemas

⚠️ Note: Default credentials (admin/admin) are for development only. Change in production!

---

🐛 Known Limitations

• Single-user mode only (multi-user support planned)
• SQLite for development (PostgreSQL recommended for production)
• No password hashing yet (plain text comparison - fix before production)
• GPU acceleration not yet implemented
• Large databases (>50K photos) may require optimization
• DeepFace model downloads on first use (can take 5-10 minutes)
• Face processing is CPU-intensive (GPU support planned for future)

---

📝 License

[Add your license here]

---

👥 Authors

PunimTag Development Team

---

🙏 Acknowledgments

• DeepFace library by Sefik Ilkin Serengil - Modern face recognition framework
• ArcFace - Additive Angular Margin Loss for Deep Face Recognition
• RetinaFace - State-of-the-art face detection
• TensorFlow, React, FastAPI, and all open-source contributors

---

📧 Support

For questions or issues:

1. Check documentation in docs/
2. See Phase 1 Checklist for implementation status
3. Review Migration Plan for roadmap

---

Made with ❤️ for photo enthusiasts

For the desktop version, see README_DESKTOP.md