ocr_engine.py

from paddleocr import PaddleOCR
import numpy as np
import re

ocr = PaddleOCR(use_angle_cls=True, lang='en', use_gpu=False)

def extract_weight_from_image(pil_img):
    try:
        img = np.array(pil_img)
        result = ocr.ocr(img, cls=True)

        all_text = []
        weight_candidates = []

        for line in result:
            for box, (text, confidence) in line:
                all_text.append(text)
                cleaned = text.lower()
                cleaned = cleaned.replace("kg", "").replace("kgs", "")
                cleaned = cleaned.replace("o", "0").replace("s", "5").replace("g", "9")
                cleaned = re.sub(r"[^\d\.]", "", cleaned)

                if re.fullmatch(r"\d{2,4}(\.\d{1,2})?", cleaned):
                    weight_candidates.append((cleaned, confidence))

        if not weight_candidates:
            return "Not detected", 0.0, "\n".join(all_text)

        best_weight, best_conf = sorted(weight_candidates, key=lambda x: -x[1])[0]
        return best_weight, round(best_conf * 100, 2), "\n".join(all_text)

    except Exception as e:
        return f"Error: {str(e)}", 0.0, "OCR failed"