Spaces:
Sleeping
Sleeping
Update ocr_engine.py
Browse files- ocr_engine.py +8 -80
ocr_engine.py
CHANGED
@@ -1,83 +1,11 @@
|
|
1 |
-
import pytesseract
|
2 |
-
import numpy as np
|
3 |
import cv2
|
4 |
-
import
|
5 |
from PIL import Image
|
6 |
-
import
|
7 |
-
|
8 |
-
# Set up logging
|
9 |
-
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
10 |
-
|
11 |
-
def preprocess_image(img):
|
12 |
-
"""Preprocess image for robust OCR."""
|
13 |
-
try:
|
14 |
-
# Convert to OpenCV format
|
15 |
-
img = np.array(img)
|
16 |
-
img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
|
17 |
-
|
18 |
-
# Convert to grayscale
|
19 |
-
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
20 |
-
|
21 |
-
# Estimate brightness for adaptive processing
|
22 |
-
brightness = np.mean(gray)
|
23 |
-
|
24 |
-
# Apply CLAHE for contrast enhancement
|
25 |
-
clahe_clip = 4.0 if brightness < 100 else 2.0
|
26 |
-
clahe = cv2.createCLAHE(clipLimit=clahe_clip, tileGridSize=(8, 8))
|
27 |
-
enhanced = clahe.apply(gray)
|
28 |
-
|
29 |
-
# Apply adaptive thresholding
|
30 |
-
block_size = max(11, min(31, int(img.shape[0] / 20) * 2 + 1))
|
31 |
-
thresh = cv2.adaptiveThreshold(
|
32 |
-
enhanced, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, block_size, 2
|
33 |
-
)
|
34 |
-
|
35 |
-
# Noise reduction
|
36 |
-
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
|
37 |
-
thresh = cv2.morphologyEx(thresh, cv2.MORPH_OPEN, kernel, iterations=1)
|
38 |
-
|
39 |
-
return thresh
|
40 |
-
except Exception as e:
|
41 |
-
logging.error(f"Preprocessing failed: {str(e)}")
|
42 |
-
return img
|
43 |
|
44 |
-
def
|
45 |
-
"""Extract weight from
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
# Preprocess image
|
52 |
-
thresh = preprocess_image(img)
|
53 |
-
|
54 |
-
# Try multiple Tesseract configurations
|
55 |
-
configs = [
|
56 |
-
r'--oem 3 --psm 7 -c tessedit_char_whitelist=0123456789.', # Single line
|
57 |
-
r'--oem 3 --psm 6 -c tessedit_char_whitelist=0123456789.' # Block of text
|
58 |
-
]
|
59 |
-
for config in configs:
|
60 |
-
text = pytesseract.image_to_string(thresh, config=config)
|
61 |
-
logging.info(f"Tesseract raw output (config {config}): {text}")
|
62 |
-
|
63 |
-
# Clean and validate text
|
64 |
-
text = re.sub(r"[^\d\.]", "", text)
|
65 |
-
if text.count('.') > 1:
|
66 |
-
text = text.replace('.', '', text.count('.') - 1)
|
67 |
-
text = text.strip('.')
|
68 |
-
if text and re.fullmatch(r"^\d*\.?\d*$", text):
|
69 |
-
text = text.lstrip('0') or '0'
|
70 |
-
confidence = 95.0 if len(text.replace('.', '')) >= 3 else 90.0
|
71 |
-
try:
|
72 |
-
weight = float(text)
|
73 |
-
if 0.001 <= weight <= 5000:
|
74 |
-
logging.info(f"Detected weight: {text} kg, Confidence: {confidence:.2f}%")
|
75 |
-
return text, confidence
|
76 |
-
except ValueError:
|
77 |
-
logging.warning(f"Invalid weight format: {text}")
|
78 |
-
|
79 |
-
logging.info("No valid weight detected.")
|
80 |
-
return "Not detected", 0.0
|
81 |
-
except Exception as e:
|
82 |
-
logging.error(f"Weight extraction failed: {str(e)}")
|
83 |
-
return "Not detected", 0.0
|
|
|
|
|
|
|
1 |
import cv2
|
2 |
+
import pytesseract
|
3 |
from PIL import Image
|
4 |
+
import numpy as np
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
+
def extract_weight(img_path):
|
7 |
+
"""Extract weight from image path using Tesseract OCR."""
|
8 |
+
img = cv2.imread(img_path)
|
9 |
+
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
10 |
+
text = pytesseract.image_to_string(gray, config='--psm 7 digits')
|
11 |
+
return ''.join(filter(lambda x: x in '0123456789.', text))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|