P&ID 기호 검출로 도면 디지털화

def run(self, tag_detections, ocr_text):
    def to_items(value):
        if value is None:
            return []
        if isinstance(value, np.ndarray):
            return to_items(value.tolist())
        if isinstance(value, (list, tuple)):
            return list(value)
        return [value]

    def clean_one_line(value):
        if value is None:
            return ""
        if isinstance(value, np.ndarray):
            return clean_one_line(value.tolist())
        if isinstance(value, dict):
            parts = []
            for v in value.values():
                s = clean_one_line(v)
                if s:
                    parts.append(s)
            return " ".join(parts)
        if isinstance(value, (list, tuple, set)):
            parts = []
            for v in value:
                s = clean_one_line(v)
                if s:
                    parts.append(s)
            return " ".join(parts)
        s = str(value)
        s = s.replace("```json", " ").replace("```", " ")
        s = " ".join(s.replace("\r", " ").replace("
", " ").replace("\t", " ").split())
        if s.lower() in ["none", "null", "[]", "{}"]:
            return ""
        return s

    raw_items = to_items(ocr_text)
    cleaned_texts = [clean_one_line(item) for item in raw_items]
    detection_count = len(tag_detections) if tag_detections is not None else 0
    if detection_count == 0:
        return {"instrument_text": []}
    output_text = []
    for i in range(detection_count):
        text = cleaned_texts[i] if i < len(cleaned_texts) else ""
        output_text.append(text if text else "unreadable")
    return {"instrument_text": output_text}