ocr-services / src /utils.py
Al-Fathir
feat : change to lightonocr
5b46afd
Raw
History Blame Contribute Delete
2.05 kB
import cv2
import numpy as np
from PIL import Image
def draw_boxes(image: np.ndarray, structured_results: list) -> Image.Image:
"""
Draws bounding boxes and text on the image when structured OCR output is available.
Bbox format: [x1, y1, x2, y2]
"""
img_bgr = image.copy()
# Define colors for different region types
colors = {
'text': (0, 255, 0), # Green
'table': (0, 0, 255), # Red
'figure': (255, 0, 0), # Blue
'title': (0, 255, 255), # Yellow
'list': (255, 0, 255), # Magenta
'unknown': (128, 128, 128) # Gray
}
for item in structured_results:
bbox = item.get('bbox', [])
region_type = item.get('type', 'unknown')
if len(bbox) == 4:
x1, y1, x2, y2 = bbox
color = colors.get(region_type, colors['unknown'])
# Draw rectangle
cv2.rectangle(img_bgr, (int(x1), int(y1)), (int(x2), int(y2)), color, 2)
# Add text label (type of region)
cv2.putText(img_bgr, region_type, (int(x1), int(y1) - 5),
cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 2)
img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
return Image.fromarray(img_rgb)
import re
def clean_ocr_markdown(raw: str | dict) -> str:
# Kalau masih dict, ambil valuenya dulu
if isinstance(raw, dict):
text = raw.get("markdown_texts") or raw.get("markdown_text", "")
else:
text = raw
# Hapus wrapper <html><body> dan </body></html> di dalam div
# Contoh: <div ...><html><body><table>...</table></body></html></div>
# → <div ...><table>...</table></div>
text = re.sub(r'<html><body>', '', text)
text = re.sub(r'</body></html>', '', text)
# Hapus div pembungkus center yang kosong atau hanya spasi
text = re.sub(r'<div[^>]*>\s*</div>', '', text)
# Rapikan multiple blank lines jadi max 2
text = re.sub(r'\n{3,}', '\n\n', text)
return text.strip()