Document Classification with OCR & QR Detection

A Flask-based web application that classifies documents (ID cards, letters, receipts, etc.) using OCR and detects QR codes in real-time via camera or uploaded images.

✨ Features

• 📄 Document Classification: Categorizes documents into KTP, Kartu Keluarga, Surat, Nota, or Lainnya
• 🔍 QR Code Detection: Real-time QR code scanning with visual overlay
• 📷 Camera Integration: Live OCR processing with stability detection
• 🖼️ Image Upload: Supports JPG/PNG uploads up to 16MB
• 🚀 Lightweight: Uses TensorFlow Lite for efficient inference

📂 Directory Structure

DCSPython
├── app.py                 # Main Flask application
├── calibrate_stability.py # Camera stability calibration tool
├── generate_qr.py         # QR code generator for testing
├── test_qr.py             # QR detection tester
├── test_camera.py         # Camera tester
├── templates/             # HTML templates
│   ├── index.html         # Homepage
│   ├── camera.html        # Camera interface
│   └── result.html        # Results page
├── static/                # Static files
│   └── style.css          # CSS styles
└── uploads/               # Temporary upload storage

🛠️ Installation

1. Clone the repository:

   git clone https://github.com/Elang-elang/DCSPython.git
   cd DCSPython

2. Install dependencies:

   pip install flask numpy opencv-python pytesseract pyzbar pillow qrcode

3. Install Tesseract OCR (required for text extraction):

   • Windows: Download from Tesseract installer
   • Mac: brew install tesseract
   • Linux: sudo apt install tesseract-ocr

🚀 Running the Application

1. Start the Flask server:

   python app.py

2. Open your browser to:

   http://localhost:5000
   

🔧 Testing Tools

• Generate test QR codes:

  python generate_qr.py

• Test camera stability:

  python calibrate_stability.py

• Test QR detection:

  python test_qr.py

📸 Usage Tips

1. For best OCR results:

   • Ensure good lighting
   • Keep the camera steady
   • Position documents flat

2. Camera mode automatically processes every 5 seconds when stable

3. Switch between normal OCR mode and dedicated QR stream mode