RoboVAI OCR & CV Core Engine 🚗🆔👤

RoboVAI OCR is a production-grade, modular Computer Vision and OCR system designed specifically for Parking Management Software Integration. It provides state-of-the-art AI capabilities for:

1. Automatic License Plate Recognition (ALPR) (Egyptian, Arabic & International plates).
2. Driver Face Verification (Face detection & authorization matching).
3. National ID & Driver License OCR (Structured 14-digit Egyptian ID parsing, DOB, governorate, expiry date, full name extraction).

🌟 Key Features

• Multi-Engine Hybrid OCR: Dual EasyOCR + PaddleOCR fallback with adaptive image preprocessing (CLAHE contrast enhancement, bilateral filtering, noise reduction, deskewing).
• High-Throughput Async Pipeline: Separates fast YOLO object tracking from heavier OCR background execution to process high-concurrency streams without dropped frames.
• Tracking Memory Cache: Caches OCR results per vehicle track_id to eliminate redundant slow OCR computations for identical cars in frame.
• Egyptian & Arabic Plate Parser: Converts raw OCR outputs into formatted plate numbers and letters (e.g., س ط أ 1 2 3).
• National ID Auto-Decoder: Decodes Egyptian 14-digit national IDs directly into date of birth, gender, century, and governorate of origin.
• Flexible Integration: Use directly as a Python SDK or deploy as a FastAPI REST API / WebSocket Stream Server with Webhooks for automated parking barrier gate control.

🚀 Quick Start & Installation

Option 1: Python SDK Installation

cd robovai_ocr
pip install -e .

Option 2: Run via CLI

# Process a single plate image
robovai-cli alpr --image path/to/car.jpg

# Process a national ID card
robovai-cli idcard --image path/to/id.jpg

# Launch API & WebSocket server
robovai-server --host 0.0.0.0 --port 8000

💻 Python SDK Usage Example

from robovai_ocr import RoboVAIEngine, Config

# Initialize engine
engine = RoboVAIEngine(device="cuda:0")

# 1. License Plate Recognition
plate_result = engine.read_license_plate("car_image.jpg")
print(plate_result)
# Output: {"plate_number": "س ط أ 1 2 3", "confidence": 0.96, "bbox": [120, 300, 450, 420]}

# 2. National ID Card Extraction
id_result = engine.extract_national_id("id_card.jpg")
print(id_result)
# Output: {
#   "national_id": "29901010100123",
#   "full_name": "احمد محمد علي",
#   "dob": "1999-01-01",
#   "gender": "Male",
#   "governorate": "Cairo"
# }

# 3. Face Matching
face_result = engine.verify_driver_face("driver_photo.jpg", db_path="data/drivers")
print(face_result)
# Output: {"match": True, "driver_id": "DRV-1029", "confidence": 0.94}

📡 REST API & WebSockets

Start the server:

robovai-server --port 8000

Open interactive docs at http://localhost:8000/docs.

REST Endpoints

• POST /api/v1/alpr: License plate detection & OCR.
• POST /api/v1/face: Driver face verification.
• POST /api/v1/id-card: National ID & driver license OCR.
• POST /api/v1/pipeline: High-throughput background queue processing.
• WS /ws/stream: Real-time RTSP/Webcam stream processing.

🏋️ Fine-Tuning & Training

To train YOLO on a custom parking plate or ID card dataset:

python training/train_yolo.py --data training/dataset_config.yaml --epochs 100