Update ocr_engine.py

ocr_engine.py

@@ -1,25 +1,51 @@
+import easyocr
+import re
 import cv2
-import pytesseract
 import numpy as np
 from PIL import Image
 
-def extract_weight_from_image(pil_img):
-    try:
-        # Convert PIL image to OpenCV
-        img = pil_img.convert("RGB")
-        img = np.array(img)
-        img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
+# Initialize EasyOCR reader (only once)
+reader = easyocr.Reader(['en'], gpu=False)
+
+def preprocess_image(image):
+    # Convert to grayscale
+    gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
+
+    # Apply thresholding (adaptive works well for 7-seg)
+    thresh = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C,
+                                   cv2.THRESH_BINARY_INV, 15, 10)
+
+    # Dilation to strengthen numbers
+    kernel = np.ones((2, 2), np.uint8)
+    dilated = cv2.dilate(thresh, kernel, iterations=1)
 
-        # Preprocess
-        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
-        blur = cv2.GaussianBlur(gray, (3, 3), 0)
+    return dilated
+
+def extract_weight_from_image(pil_image):
+    try:
+        # Convert PIL to OpenCV
+        image = np.array(pil_image.convert("RGB"))
+        processed = preprocess_image(image)
 
         # OCR
-        text = pytesseract.image_to_string(blur, config='--psm 7 digits')
-        weight = ''.join(filter(lambda c: c in '0123456789.', text))
+        result = reader.readtext(processed)
+
+        # Filter and extract digits like weight (e.g., 75.5)
+        weight = None
+        confidence = 0.0
+        for detection in result:
+            text = detection[1]
+            conf = detection[2]
+            match = re.search(r"\d{2,4}(\.\d{1,2})?", text)  # match 2-4 digit decimal
+            if match:
+                weight = match.group()
+                confidence = conf
+                break
+
+        if weight:
+            return weight, round(confidence * 100, 2)
+        else:
+            return "No weight detected", 0.0
 
-        confidence = 95  # Replace with real confidence logic if needed
-        return weight.strip(), confidence
     except Exception as e:
-        print(f"OCR error: {e}")
-        return "", 0
+        return f"Error: {str(e)}", 0.0