Update ocr_engine.py

ocr_engine.py

@@ -1,55 +1,52 @@
-import numpy as np
 import re
 import cv2
+import numpy as np
 from PIL import Image
-import easyocr
+from paddleocr import PaddleOCR
 
-# ✅ Initialize EasyOCR Reader once
-reader = easyocr.Reader(['en'], gpu=False)
+# ✅ Initialize PaddleOCR once
+ocr = PaddleOCR(use_angle_cls=False, lang='en', show_log=False)
 
 def preprocess_image(image):
     """
-    Convert to grayscale and apply adaptive thresholding
-    to enhance contrast for digital scale OCR.
+    Convert to grayscale and enhance contrast to help OCR.
     """
     gray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
-    thresh = cv2.adaptiveThreshold(
+    processed = cv2.adaptiveThreshold(
         gray, 255,
-        cv2.ADAPTIVE_THRESH_MEAN_C,
+        cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
         cv2.THRESH_BINARY_INV,
-        11, 10
+        11, 2
     )
-    return thresh
+    return processed
 
 def extract_weight_from_image(pil_image):
     try:
-        # ✅ Convert PIL image to OpenCV format
+        # ✅ Convert to OpenCV format
         image = np.array(pil_image.convert("RGB"))
 
         # ✅ Preprocess image
         processed = preprocess_image(image)
 
-        # ✅ Optional: Save debug image for troubleshooting
+        # Optional: Save debug image
         debug_path = "debug_processed_image.png"
         Image.fromarray(processed).save(debug_path)
-        print(f"[DEBUG] Preprocessed image saved to: {debug_path}")
+        print(f"[DEBUG] Saved preprocessed image to {debug_path}")
 
-        # ✅ Perform OCR using EasyOCR
-        result = reader.readtext(processed)
+        # ✅ Run OCR on original (RGB) image (not preprocessed)
+        result = ocr.ocr(image, cls=False)
 
-        print("🔍 OCR Results:")
-        for detection in result:
-            print(f"  • Text: '{detection[1]}' | Confidence: {detection[2]*100:.2f}%")
+        print("🔍 PaddleOCR Results:")
+        for line in result[0]:
+            text = line[1][0]
+            conf = line[1][1]
+            print(f"  • Text: '{text}' | Confidence: {conf*100:.2f}%")
 
-        # ✅ Extract first matching numeric value
-        for detection in result:
-            text = detection[1].replace(",", ".")  # normalize decimal
-            conf = detection[2]
+            # Try to extract number like 53.25 or 100
             match = re.search(r"\b\d{1,4}(\.\d{1,2})?\b", text)
             if match:
                 return match.group(), round(conf * 100, 2)
 
-        # ❌ No weight found
         return "No weight detected", 0.0
 
     except Exception as e: