Update ocr_engine.py

ocr_engine.py

@@ -5,41 +5,35 @@ from PIL import Image
 
 def extract_weight_from_image(pil_img):
     try:
-        # Convert PIL image to OpenCV format
-        img = pil_img.convert("RGB")
-        img_np = np.array(img)
-        img_cv = cv2.cvtColor(img_np, cv2.COLOR_RGB2BGR)
-        print("🧼 Image converted to OpenCV format.")
-
-        # Convert to grayscale
-        gray = cv2.cvtColor(img_cv, cv2.COLOR_BGR2GRAY)
-
-        # Adaptive Thresholding for 7-segment LCD
-        processed = cv2.adaptiveThreshold(
-            gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, 15, 10
-        )
+        # Step 1: Convert PIL to OpenCV
+        img = pil_img.convert("L")  # grayscale
+        img = np.array(img)
+
+        # Step 2: Resize image for better OCR accuracy
+        img = cv2.resize(img, None, fx=2, fy=2, interpolation=cv2.INTER_CUBIC)
 
-        # Resize to enhance small text
-        resized = cv2.resize(processed, None, fx=2, fy=2, interpolation=cv2.INTER_LINEAR)
+        # Step 3: Apply Gaussian Blur to remove noise
+        blur = cv2.GaussianBlur(img, (5, 5), 0)
 
-        # Optional: Apply dilation to enhance contours
-        kernel = np.ones((2, 2), np.uint8)
-        dilated = cv2.dilate(resized, kernel, iterations=1)
+        # Step 4: Apply Adaptive Thresholding
+        thresh = cv2.adaptiveThreshold(
+            blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
+            cv2.THRESH_BINARY_INV, 11, 2
+        )
 
-        # OCR config tuned for digits
-        config = r'--oem 3 --psm 7 -c tessedit_char_whitelist=0123456789.'
+        # Step 5: OCR Config - digits only
+        config = r'--oem 3 --psm 6 -c tessedit_char_whitelist=0123456789.'
 
-        # Run OCR
-        text = pytesseract.image_to_string(dilated, config=config)
-        print("🔍 RAW OCR OUTPUT:", repr(text))
+        # Step 6: Run OCR
+        text = pytesseract.image_to_string(thresh, config=config)
+        print("🔍 OCR RAW OUTPUT:", repr(text))  # view this in Hugging Face logs
 
-        # Clean the text
-        weight = ''.join(c for c in text if c in '0123456789.')
-        weight = weight.strip()
+        # Step 7: Extract numbers
+        weight = ''.join(filter(lambda c: c in '0123456789.', text))
 
         confidence = 95 if weight else 0
-        return weight, confidence
+        return weight.strip(), confidence
 
     except Exception as e:
-        print("❌ OCR Error:", str(e))
-        return "", 0
+        print("❌ OCR Exception:", str(e))
+        return "", 0