Update ocr_engine.py

ocr_engine.py

@@ -1,33 +1,42 @@
-from paddleocr import PaddleOCR
+import easyocr
 import numpy as np
+import cv2
 import re
 
-ocr = PaddleOCR(use_angle_cls=True, lang='en', use_gpu=False)
+reader = easyocr.Reader(['en'], gpu=False)
 
 def extract_weight_from_image(pil_img):
     try:
         img = np.array(pil_img)
-        result = ocr.ocr(img, cls=True)
 
-        all_text = []
+        # Resize and grayscale
+        img = cv2.resize(img, None, fx=4, fy=4, interpolation=cv2.INTER_LINEAR)
+        gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
+        gray = cv2.bilateralFilter(gray, 11, 17, 17)
+        _, thresh = cv2.threshold(gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+
+        results = reader.readtext(thresh)
+
+        ocr_raw_texts = []
         weight_candidates = []
 
-        for line in result:
-            for box, (text, confidence) in line:
-                all_text.append(text)
-                cleaned = text.lower()
-                cleaned = cleaned.replace("kg", "").replace("kgs", "")
-                cleaned = cleaned.replace("o", "0").replace("s", "5").replace("g", "9")
-                cleaned = re.sub(r"[^\d\.]", "", cleaned)
+        for _, text, conf in results:
+            ocr_raw_texts.append(text)
+            t = text.lower()
+            t = t.replace("kg", "").replace("kgs", "")
+            t = t.replace("o", "0").replace("O", "0")
+            t = t.replace("s", "5").replace("S", "5")
+            t = t.replace("g", "9").replace("G", "6")
+            t = re.sub(r"[^\d\.]", "", t)
 
-                if re.fullmatch(r"\d{2,4}(\.\d{1,2})?", cleaned):
-                    weight_candidates.append((cleaned, confidence))
+            if re.fullmatch(r"\d{2,4}(\.\d{1,2})?", t):
+                weight_candidates.append((t, conf))
 
         if not weight_candidates:
-            return "Not detected", 0.0, "\n".join(all_text)
+            return "Not detected", 0.0, "\n".join(ocr_raw_texts)
 
         best_weight, best_conf = sorted(weight_candidates, key=lambda x: -x[1])[0]
-        return best_weight, round(best_conf * 100, 2), "\n".join(all_text)
+        return best_weight, round(best_conf * 100, 2), "\n".join(ocr_raw_texts)
 
     except Exception as e:
         return f"Error: {str(e)}", 0.0, "OCR failed"