Update ocr_engine.py

ocr_engine.py

@@ -5,29 +5,31 @@ from PIL import Image
 
 def extract_weight_from_image(pil_img):
     try:
-        # Convert to OpenCV image
+        # Convert PIL to OpenCV
         img = pil_img.convert("RGB")
         img = np.array(img)
         img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
 
-        # Crop or resize if needed (optional based on display layout)
-
-        # Convert to grayscale and enhance contrast
+        # Convert to grayscale
         gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
-        _, thresh = cv2.threshold(gray, 150, 255, cv2.THRESH_BINARY_INV)
 
-        # Optional: Resize for better OCR (helps with small digits)
-        resized = cv2.resize(thresh, None, fx=2, fy=2, interpolation=cv2.INTER_LINEAR)
+        # Thresholding to highlight digits
+        _, binary = cv2.threshold(gray, 120, 255, cv2.THRESH_BINARY_INV)
+
+        # Resize for better OCR
+        resized = cv2.resize(binary, None, fx=2, fy=2, interpolation=cv2.INTER_CUBIC)
+
+        # Run OCR with digit whitelist
+        config = r'--oem 3 --psm 6 -c tessedit_char_whitelist=0123456789.'
+        raw_text = pytesseract.image_to_string(resized, config=config)
 
-        # Apply OCR with proper config for digits
-        custom_config = r'--oem 3 --psm 6 -c tessedit_char_whitelist=0123456789.'
-        text = pytesseract.image_to_string(resized, config=custom_config)
+        print("🔍 OCR Raw Output:", repr(raw_text))  # Show in Hugging Face logs
 
-        # Filter out non-numeric parts
-        weight = ''.join(filter(lambda c: c in '0123456789.', text))
+        # Filter for digits only
+        weight = ''.join(filter(lambda c: c in '0123456789.', raw_text))
         confidence = 95 if weight else 0
         return weight.strip(), confidence
 
     except Exception as e:
-        print("OCR Exception:", str(e))
+        print("❌ OCR Error:", str(e))
         return "", 0