Update ocr_engine.py

ocr_engine.py

@@ -5,38 +5,51 @@ import re
 
 reader = easyocr.Reader(['en'], gpu=False)
 
+def enhance_image(img):
+    # Resize to make text clearer
+    img = cv2.resize(img, None, fx=4, fy=4, interpolation=cv2.INTER_CUBIC)
+
+    # Convert to gray
+    gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
+
+    # Denoise
+    gray = cv2.fastNlMeansDenoising(gray, h=15)
+
+    # Sharpen
+    kernel = np.array([[0, -1, 0],
+                       [-1, 5,-1],
+                       [0, -1, 0]])
+    sharp = cv2.filter2D(gray, -1, kernel)
+
+    # Contrast enhancement (CLAHE)
+    clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
+    contrast = clahe.apply(sharp)
+
+    return contrast
+
 def extract_weight_from_image(pil_img):
     try:
         img = np.array(pil_img)
+        enhanced = enhance_image(img)
 
-        # 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 = []
+        results = reader.readtext(enhanced)
+        ocr_texts = [text for _, text, _ in results]
         weight_candidates = []
 
         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)
+            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)
 
-            if re.fullmatch(r"\d{2,4}(\.\d{1,2})?", t):
-                weight_candidates.append((t, conf))
+            if re.fullmatch(r"\d{2,4}(\.\d{1,2})?", cleaned):
+                weight_candidates.append((cleaned, conf))
 
         if not weight_candidates:
-            return "Not detected", 0.0, "\n".join(ocr_raw_texts)
+            return "Not detected", 0.0, "\n".join(ocr_texts)
 
         best_weight, best_conf = sorted(weight_candidates, key=lambda x: -x[1])[0]
-        return best_weight, round(best_conf * 100, 2), "\n".join(ocr_raw_texts)
+        return best_weight, round(best_conf * 100, 2), "\n".join(ocr_texts)
 
     except Exception as e:
         return f"Error: {str(e)}", 0.0, "OCR failed"