Update ocr_engine.py

ocr_engine.py

@@ -2,9 +2,9 @@ from transformers import TrOCRProcessor, VisionEncoderDecoderModel
 from PIL import Image
 import re
 
-# Load smaller/faster model
-processor = TrOCRProcessor.from_pretrained("microsoft/trocr-small-printed")
-model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-small-printed")
+# Load lightweight, fast, public model
+processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")
+model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-handwritten")
 
 def extract_weight(image: Image.Image) -> str:
     image = image.convert("RGB")
@@ -12,10 +12,12 @@ def extract_weight(image: Image.Image) -> str:
     generated_ids = model.generate(pixel_values, max_length=32)
     full_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
 
+    # Extract number
     cleaned = full_text.lower().replace(" ", "")
     match = re.search(r"(\d+(\.\d+)?)", cleaned)
     weight = match.group(1) if match else None
 
+    # Detect unit
     if any(u in cleaned for u in ["kg", "kgs", "kilogram", "kilo"]):
         unit = "kg"
     elif any(u in cleaned for u in ["g", "gram", "grams"]):