Update ocr_engine.py

ocr_engine.py

@@ -8,7 +8,6 @@ processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")
 model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-handwritten")
 
 def clean_ocr_text(text):
-    # Fix common OCR misreads
     text = text.replace(",", ".").replace("s", "5").replace("o", "0").replace("O", "0")
     return re.sub(r"[^\d.kg]", "", text.lower())
 
@@ -21,41 +20,39 @@ def restore_decimal(text):
 
 def extract_unit_from_text(raw_text):
     raw_text = raw_text.lower()
-    if "kg" in raw_text:
+    if "kg" in raw_text or "kgs" in raw_text or "k9" in raw_text:
         return "kg"
-    elif "g" in raw_text:
+    elif re.search(r'\dg', raw_text) or "gram" in raw_text:
         return "g"
-    return "g"  # fallback if no unit
+    return None
 
 def extract_weight(image):
     try:
-        # Resize & sharpen image
         image = image.resize((image.width * 2, image.height * 2), Image.BICUBIC)
         image = image.filter(ImageFilter.SHARPEN)
 
-        # OCR inference
         pixel_values = processor(images=image, return_tensors="pt").pixel_values
         generated_ids = model.generate(pixel_values)
         raw_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0].strip()
 
         cleaned = clean_ocr_text(raw_text)
 
-        # Case 1: Match decimal with unit
         match = re.search(r"(\d{1,3}\.\d{1,3})\s*(kg|g)?", cleaned)
         if match:
-            return f"{match.group(1)} {match.group(2) or ''}".strip(), raw_text
+            weight = match.group(1)
+            unit = match.group(2) or extract_unit_from_text(raw_text)
+            return f"{weight} {unit or ''}".strip(), raw_text
 
-        # Case 2: Large number fallback like 53255 → 52.255
         match = re.search(r"\d{4,5}", cleaned)
         if match:
             decimal_fixed = restore_decimal(match.group())
             unit = extract_unit_from_text(raw_text)
-            return f"{decimal_fixed} {unit}", raw_text
+            return f"{decimal_fixed} {unit or ''}".strip(), raw_text
 
-        # Final fallback: plain number
         match = re.search(r"\d+", cleaned)
         if match:
-            return f"{match.group()} g", raw_text
+            unit = extract_unit_from_text(raw_text)
+            return f"{match.group()} {unit or ''}".strip(), raw_text
 
         return "Error: No valid weight found", raw_text