Sanjayraju30 commited on
Commit
ee1d691
·
verified ·
1 Parent(s): 24c2f9b

Update ocr_engine.py

Browse files
Files changed (1) hide show
  1. ocr_engine.py +28 -14
ocr_engine.py CHANGED
@@ -1,29 +1,43 @@
1
- import pytesseract
2
  import numpy as np
3
  import cv2
4
  import re
5
 
 
 
 
6
  def extract_weight_from_image(pil_img):
7
  try:
8
- # Convert PIL to OpenCV format
9
  img = np.array(pil_img)
10
 
11
- # Resize and convert to grayscale
12
- img = cv2.resize(img, None, fx=2, fy=2, interpolation=cv2.INTER_CUBIC)
13
  gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
 
 
 
 
 
 
 
 
 
 
 
 
 
14
 
15
- # Thresholding for clarity
16
- _, thresh = cv2.threshold(gray, 150, 255, cv2.THRESH_BINARY_INV)
 
17
 
18
- # Run OCR
19
- text = pytesseract.image_to_string(thresh)
20
- print("OCR TEXT:", text)
21
 
22
- # Search for weight pattern
23
- match = re.search(r"\b\d{2,4}(\.\d{1,2})?\b", text)
24
- if match:
25
- return match.group(), 99.0
26
- return "Not detected", 0.0
27
 
28
  except Exception as e:
29
  return f"Error: {str(e)}", 0.0
 
1
+ import easyocr
2
  import numpy as np
3
  import cv2
4
  import re
5
 
6
+ # Initialize OCR reader once
7
+ reader = easyocr.Reader(['en'], gpu=False)
8
+
9
  def extract_weight_from_image(pil_img):
10
  try:
11
+ # Convert image to NumPy format
12
  img = np.array(pil_img)
13
 
14
+ # Resize and preprocess
15
+ img = cv2.resize(img, None, fx=3.5, fy=3.5, interpolation=cv2.INTER_LINEAR)
16
  gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
17
+ gray = cv2.bilateralFilter(gray, 11, 17, 17)
18
+ _, thresh = cv2.threshold(gray, 120, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)
19
+
20
+ # OCR
21
+ results = reader.readtext(thresh)
22
+
23
+ # Debug
24
+ print("OCR Results:", results)
25
+
26
+ weight_candidates = []
27
+ for _, text, conf in results:
28
+ clean = text.lower().replace("kg", "").strip()
29
+ clean = clean.replace("o", "0").replace("O", "0") # fix OCR misreads
30
 
31
+ # Match weights like 86, 85.5, 102.3
32
+ if re.fullmatch(r"\d{2,4}(\.\d{1,2})?", clean):
33
+ weight_candidates.append((clean, conf))
34
 
35
+ if not weight_candidates:
36
+ return "Not detected", 0.0
 
37
 
38
+ # Return best candidate
39
+ best_weight, best_conf = sorted(weight_candidates, key=lambda x: -x[1])[0]
40
+ return best_weight, round(best_conf * 100, 2)
 
 
41
 
42
  except Exception as e:
43
  return f"Error: {str(e)}", 0.0