| Title |
Automated Extraction of Fire Protection System Design Information from Architectural Drawings Using Artificial Intelligence |
| Authors |
Sang Hun Yeon ; Chul Ho Kim ; Kye-Won Park ; Doo Chan Choi ; Yonggoel Jo ; Jong Min Choi ; Kwang Ho Lee |
| DOI |
https://doi.org/10.6110/KJACR.2023.35.7.331 |
| Keywords |
자동 인식; 표 감지; 광학 문자 인식; 문자 추출; 인공지능 Automated recognition; Table detector; Optical character recognition; Character extraction; Artificial intelligence |
| Abstract |
This study aimed to develop an artificial intelligence-based fire protection system design engineering solution. We developed an automatic fire protection system design element extraction algorithm from architectural overview tables using various artificial intelligence libraries. The deep learning-based libraries used in this study were TableNet, OpenCV, and EasyOCR. Levenshtein distance was also used to check the similarity of the characters. Approximately 1,000 tables and the Marmot dataset, which is open-source training data, were trained for the development of the table detection model. The performance metrics used for performance evaluation were recall, precision, and F1 score. The final model selected had the highest F1 score of 0.63. The results confirmed that fire protection system design elements could be efficiently extracted through the developed algorithm. We expect various positive effects of this newly developed algorithm, such as improvements in engineers’ working productivity, functional suitability of designs, and reliability. |