import cv2

import numpy as np

from urllib.request import urlopen

from meikiocr import MeikiOCR

# --- 1. Load Assets ---

# load from url...

IMAGE_URL = "https://huggingface.co/spaces/rtr46/meikiocr/resolve/main/example.jpg"

print(f"--- Downloading sample image from {IMAGE_URL} ---")

with urlopen(IMAGE_URL) as resp:

image_bytes = np.asarray(bytearray(resp.read()), dtype="uint8") # Read image data into a buffer

image = cv2.imdecode(image_bytes, cv2.IMREAD_COLOR) # Decode buffer into an OpenCV image

# ... or load from disk

# image = cv2.imread("my_input_image.jpg", cv2.IMREAD_COLOR)

print("--- Initializing MeikiOCR (models will be downloaded on first run) ---")

ocr = MeikiOCR() # Initialize the OCR pipeline

# --- 2. Run OCR ---

print("--- Running OCR on the image ---")

results = ocr.run_ocr(image, punct_conf_factor=0.2) # Run the full OCR pipeline

# --- 3. Display and Save Results ---

print("\n--- Recognized Text ---")

full_text = '\n'.join([line['text'] for line in results if line['text']])

print(full_text)

print("-----------------------")

vis_image = image.copy() # Create a copy of the image for drawing

for line in results:

for char_info in line['chars']:

x1, y1, x2, y2 = char_info['bbox'] # Get character bounding box

cv2.rectangle(vis_image, (x1, y1), (x2, y2), (0, 255, 0), 1) # Draw a green box

output_path = "demo_result.jpg"

cv2.imwrite(output_path, vis_image) # Save the image with boxes

print(f"\n--- Saved visualization with character boxes to {output_path} ---")