Update ocr_engine.py

ocr_engine.py

@@ -1,9 +1,9 @@
 from transformers import DonutProcessor, VisionEncoderDecoderModel
 from PIL import Image
-import torch
 import re
+import torch
 
-# Load model
+# Load processor + model
 processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")
 model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")
 
@@ -11,21 +11,19 @@ def extract_weight(image: Image.Image) -> str:
     image = image.convert("RGB")
     pixel_values = processor(image, return_tensors="pt").pixel_values
 
-    # Generate output
+    # Generate text prediction
     outputs = model.generate(pixel_values, max_length=512)
     decoded = processor.batch_decode(outputs, skip_special_tokens=True)[0]
 
-    print("OCR Output:", decoded)  # Optional for debug
-
-    # Extract number
-    match = re.search(r"(\d+(\.\d+)?)", decoded)
+    # Clean & extract weight
+    cleaned = decoded.lower().replace(" ", "")
+    match = re.search(r"(\d+(\.\d+)?)", cleaned)
     weight = match.group(1) if match else None
 
     # Detect unit
-    text_lower = decoded.lower().replace(" ", "")
-    if any(u in text_lower for u in ["kg", "kgs", "kilogram", "kilo"]):
+    if any(u in cleaned for u in ["kg", "kgs", "kilogram", "kilo"]):
         unit = "kg"
-    elif any(u in text_lower for u in ["g", "gram", "grams"]):
+    elif any(u in cleaned for u in ["g", "gram", "grams"]):
         unit = "grams"
     else:
         unit = "kg" if weight and float(weight) >= 5 else "grams"