pip install ultralytics easyocr opencv-python pillow torch torchvision huggingface_hub

import cv2
import numpy as np
from ultralytics import YOLO
import easyocr
from PIL import Image
from huggingface_hub import hf_hub_download
import warnings

# Suppress warnings
warnings.filterwarnings('ignore')

# Download models from HuggingFace
print("Downloading model from HuggingFace...")
model_path = hf_hub_download(repo_id="0xnu/european-license-plate-recognition", filename="model.onnx")
config_path = hf_hub_download(repo_id="0xnu/european-license-plate-recognition", filename="config.json")

# Load models with explicit task specification
yolo_model = YOLO(model_path, task='detect')
ocr_reader = easyocr.Reader(['en', 'de', 'fr', 'es', 'it', 'nl'], gpu=False, verbose=False)

# Process image
def recognize_license_plate(image_path):
   # Load image
   image = cv2.imread(image_path)
   image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
   
   # Detect license plates
   results = yolo_model(image_rgb, conf=0.5, verbose=False)
   
   plates = []
   for result in results:
       boxes = result.boxes
       if boxes is not None:
           for box in boxes:
               # Get coordinates
               x1, y1, x2, y2 = box.xyxy[0].cpu().numpy()
               
               # Crop plate
               plate_crop = image_rgb[int(y1):int(y2), int(x1):int(x2)]
               
               # Extract text
               ocr_results = ocr_reader.readtext(plate_crop)
               if ocr_results:
                   text = ocr_results[0][1]
                   confidence = float(ocr_results[0][2])  # Convert to native Python float
                   plates.append({'text': text, 'confidence': confidence})
   
   return plates

# Usage Example
results = recognize_license_plate('sample_car_with_license.jpeg')
print(results)