Journal of the Architectural Institute of Korea (대한건축학회논문집)

Architectural Institute of Korea (대한건축학회)
권호 표지
DOI QR코드

DOI QR Code

Material Database Extraction Through Deep Learning From Bill of Quantity

딥러닝을 활용한 건축내역서 자재 데이터베이스 추출

  • Miao, Xu (School of Architecture, Kyungil University) ;
  • Eom, Shin-Jo (School of Architecture, Kyungil University)
  • Received : 2023.10.29
  • Accepted : 2023.11.27
  • Published : 2023.12.31
⟨ Previous Next ⟩

Abstract

In construction projects, optimizing material selection is crucial, as over half of the construction cost is allocated to materials. To achieve this, an integrated material information system becomes essential. Creating an efficient material list requires significant investment in manpower and time to register and manage diverse material information. This study introduces a system developed through deep learning-based intelligent material extraction. The system builds a database of building material information from real projects, utilizing a classifier trained with standard construction codes using the FastText method and LSTM model. Through experiments on 40 buildings, the system demonstrated an 86% accuracy rate. The resulting building material information serves as a foundational resource for future applications such as artificial intelligence-based automation of design economic evaluation and design safety assessment.

Keywords

Acknowledgement

이 연구는 2020년도 한국연구재단 연구비 지원에 의한 결과의 일부임. 과제번호:2020R1I1A3073663

References

  1. Gu, H., Seo, J., & Choo, S. (2019). A Development of Facade Dataset Construction Technology Using Deep Learning-based Automatic Image Labeling, Journal of The Architectural Institute of Korea, Planning & Design, 35(12), 43-53.
  2. Ha, Y., Yu, J., & Kim, C. (2009). The material management process model of material characteristic for material administration of physical distribution, Korea Journal of Construction Engineering and Management, 10(1), 16~25.
  3. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory, Neural computation, 9(8), pp. 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735
  4. Joulin, A., Grave, E., Bojanowski, P., & Mikolov, T. (2016). Bag of tricks for efficient text classification, Association for Computational Linguistics, Proceedings of the 15th Conference of the European Chapter of the Association for Computational, 2, 427-431.
  5. Kwon, O., Han, C., & Lee, H. (2007). A Research on how to turn informationization Database of construction materials to practical use, Korea Institute of Construction Engineering and Management, Create Blue Ocean, 747-750.
  6. Lee, Y., & Kim, K. (2020). Experimental Study on the Expansion of the Short-term Prediction Range of Rebar Prices Using Deep Learning, Journal of The Architectural Institute of Korea, 36(12), 265-272. https://doi.org/10.5659/JAIK.2020.36.12.265
  7. Lee, J., Park, J., & Yoon, H. (2020). Automatic Classification of Bridge Component based on Deep Learning, Journal of The Korean Society of Civil Engineers, 40(2), 265-272.
  8. Mikolov, T., Chen K., Corrado G., & Dean J. (2013). Efficient estimation of word representations in vector space, 1st International Conference on Learning Representations ICLR, Scottsdale, Arizona, USA, Workshop Track Proceedings arXiv:1301.3781
  9. Park, H., & Kook, K. (2014). A hierarchical model of housing choice, Journal of The Architectural Institute of Korea, Structure & Construction, 30(5), 109-116. https://doi.org/10.5659/JAIK_SC.2014.30.5.109
  10. Song, J., Han, C., & Ju, K. (2011). A Study on Operational Plan of the System for Construction Material Information, Society of Design Convergence, 10(2), 63-75.
  11. Schuster, M., & Paliwal, K.K. (1997). Bidirectional recurrent neural networks, IEEE Trans Signal Process, 45, pp. 2673-2681.  https://doi.org/10.1109/78.650093