ocr_engine.py

from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image
import re

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

def extract_weight(image: Image.Image) -> str:
    image = image.convert("RGB")
    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]

    # Lowercase text but don't strip spacing before kg detection
    full_text_lower = full_text.lower()

    # Detect unit
    if "kg" in full_text_lower.replace(" ", ""):
        unit = "kg"
    elif "g" in full_text_lower.replace(" ", "") or "gram" in full_text_lower:
        unit = "grams"
    else:
        unit = "grams"  # default

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