moamen-essmat/egypt-ocr

# 🚗 Egyptian License Plate Detection & OCR System A complete system for detecting and reading Egyptian license plates using YOLOv11 for detection and PaddleOCR for Arabic text recognition. ## Project Structure ``` Egyptian-Plate-OCR/ ├── train_detection.ipynb # Training notebook (download dataset + train) ├── inference.ipynb # Test on images/video ├── detect.py # CLI inference script ├── plate_ocr.py # OCR utilities ├── requirements.txt # Dependencies ├── data/ # Dataset (after download) ├── test_images/ # Test images/videos └── runs/ # Training runs ``` ## Quick Start ### 1. Install Dependencies ```bash cd C:\Users\Admin\Egyptian-Plate-OCR pip install -r requirements.txt ``` ### 2. Get Dataset 1. Create free account at app.roboflow.com 2. Go to: app.roboflow.com 3. Click "Download" → select "YOLOv11" format 4. Copy your Roboflow API key ### 3. Train Model Open `train_detection.ipynb` in VS Code: - Add your Roboflow API key - Run all cells - Best model saved to `best.pt` ### 4. Run Inference ```bash # Single image python detect.py --source test_images/test.jpg # Video python detect.py --source test_images/test.mp4 ``` Or use `inference.ipynb` for interactive testing. ## Egyptian Plate Format Egyptian plates typically contain: - **2-3 digits** (Arabic numerals: ١٢٣) - **Arabic letter** (governor code) - **2-3 digits** (serial number) Example: `١٢٣ هـ ٤٥٦` → `123H456` ## Model Performance | Metric | Score | |-----------|--------| | Precision | 99.94% | | Recall | 99.49% | | mAP@50 | 99.50% | ## Dependencies - Ultralytics YOLOv11 - PaddleOCR (Arabic support) - OpenCV - NumPy - Matplotlib