ocr_engine.py

from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image

# Load OCR model once
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-stage1")
model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-stage1")

def extract_weight(image: Image.Image) -> str:
    # Ensure image is in RGB
    image = image.convert("RGB")

    # Process with Hugging Face OCR
    pixel_values = processor(images=image, return_tensors="pt").pixel_values
    generated_ids = model.generate(pixel_values)
    full_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

    # Normalize text
    full_text_cleaned = full_text.lower().replace(" ", "")

    # Detect unit
    if "kg" in full_text_cleaned:
        unit = "kg"
    elif "g" in full_text_cleaned or "gram" in full_text_cleaned:
        unit = "grams"
    else:
        unit = "grams"  # default to grams if not clear

    # Extract number (includes decimals)
    import re
    match = re.search(r"(\d+(\.\d+)?)", full_text_cleaned)
    if match:
        weight = match.group(1)
        return f"{weight} {unit}"
    else:
        return "No valid weight detected"