Update ocr_engine.py

ocr_engine.py

@@ -1,32 +1,28 @@
-import easyocr
+from mmocr.apis import MMOCRInferencer
 import numpy as np
 import cv2
 import re
+from PIL import Image
 
-# Load EasyOCR reader
-reader = easyocr.Reader(['en'], gpu=False)
+# Initialize MMOCR
+ocr = MMOCRInferencer(det='DBNet', recog='SAR', device='cpu')  # or 'cuda' if GPU available
 
 def extract_weight_from_image(pil_img):
     try:
-        img = np.array(pil_img)
+        # Convert PIL to OpenCV image (BGR)
+        img = np.array(pil_img.convert("RGB"))[:, :, ::-1]
 
-        # Resize very large images
-        max_dim = 1000
-        height, width = img.shape[:2]
-        if max(height, width) > max_dim:
-            scale = max_dim / max(height, width)
-            img = cv2.resize(img, None, fx=scale, fy=scale, interpolation=cv2.INTER_AREA)
-
-        # OCR recognition
-        results = reader.readtext(img)
-        print("DEBUG OCR RESULTS:", results)
+        # Run MMOCR inference
+        result = ocr(img)
 
         raw_texts = []
         weight_candidates = []
         fallback_weight = None
         fallback_conf = 0.0
 
-        for _, (text, conf) in results:
+        for item in result['predictions'][0]:
+            text = item['text']
+            conf = item.get('score', 0.8)  # Fallback confidence
             original = text
             cleaned = text.lower().strip()
 
@@ -40,16 +36,13 @@ def extract_weight_from_image(pil_img):
 
             raw_texts.append(f"{original} → {cleaned} (conf: {round(conf, 2)})")
 
-            # Save fallback if no match later
             if cleaned and cleaned.replace(".", "").isdigit() and not fallback_weight:
                 fallback_weight = cleaned
                 fallback_conf = conf
 
-            # Match proper weight format: 75.02, 97.2, 105
             if cleaned.count(".") <= 1 and re.fullmatch(r"\d{2,4}(\.\d{1,3})?", cleaned):
                 weight_candidates.append((cleaned, conf))
 
-        # Choose best candidate
         if weight_candidates:
             best_weight, best_conf = sorted(weight_candidates, key=lambda x: -x[1])[0]
         elif fallback_weight:
@@ -57,7 +50,6 @@ def extract_weight_from_image(pil_img):
         else:
             return "Not detected", 0.0, "\n".join(raw_texts)
 
-        # Strip unnecessary leading zeros
         if "." in best_weight:
             int_part, dec_part = best_weight.split(".")
             int_part = int_part.lstrip("0") or "0"