Published Papers

Volume 32 - Issue 4 (Award Issue)
Award Papers

YOSO (You Only Scan Once): A Large-Small Model Co-Adapter Framework for AI–Human Collaborative Scan-to-BIM Automation in Wastewater Infrastructure

Winson TC Leung, CH Wong, Senna CS Ng, Simon YO Lai, Carry PS Cheung, CY Lam, YM Qin and LF Ren
Pages: 1-11Published: 31 Dec 2025
DOI: 10.33430/V32N4THIE-2025-0040
Cite thisHide

Leung TC, Wong CH, Ng CS, Lai YO, Cheung PS, Lam CY, Qin YM and Ren LF, YOSO (You Only Scan Once): A Large-Small Model Co-Adapter Framework for AI–Human Collaborative Scan-to-BIM Automation in Wastewater Infrastructure, HKIE Transactions, Vol. 32, No. 4 (Award Issue), Article THIE-2025-0040.R1, 2025, 10.33430/V32N4THIE-2025-0040

 Copy

Abstract:

Hong Kong’s ageing wastewater infrastructure faces escalating climate risks, necessitating efficient asset management and swift upgrades through Modular Integrated Construction (MiC) and Multi-trade integrated Mechanical, Electrical and Plumbing (MiMEP) workflows. These efforts demand precise as-built BIM, yet traditional Scan-to-BIM methods remain hindered by labour-intensive segmentation, outdated semantic data and fragmented toolsets. This paper presents YOSO (You Only Scan Once), an AI-assisted framework built on a Large-Small Model Co-Adapter pipeline to automate Scanto-BIM conversion. The Co-Adapter synergises a small model in a lightweight edge device, MindPalace Pocket, for realtime spatial data capture and a large model in a cloud-based DBSCAN-driven workflow for clustering and BIM generation. A confidence-driven interface routes low-certainty detections to engineers, reducing manual input by 80% while ensuring accuracy. Novel metrics—Intersection-to-Manual Ratio (IMR) for component prefabrication and Euclidean Distance (ED) for clash-free installation—are proposed for evaluation. Validated at Ting Kok Road Sewage Pumping Station No.8 using light tubes (representing standardised components) and gate valves (representing irregular geometries), YOSO achieves 99.9% and 84.6% average IMR, with minor average ED of 0.160 m and 0.116 m, respectively. By bridging edge-cloud AI with human expertise, YOSO directly supports digitalisation goals, offering a scalable blueprint for modernising ageing infrastructure globally.

Keywords:

Adaptive segmentation, AI–human collaboration, Building Information Modelling (BIM), Edge-cloud computing, Infrastructure resilience, Scan-to-BIM

Reference List:

1. Jiang, F., Ma, L., Broyd, T. & Chen, K. (2021), ‘Digital twin and its implementations in the civil engineering
sector’, Automation in Construction 130, 103838.
2. Justo Dom´ınguez, A., Soil´an, M., S´anchez Rodr´ıguez, A. & Riveiro, B. (2021), ‘Scan-to-bim for the infrastructure domain: Generation of ifccompliant models of road infrastructure assets and semantics using 3d point cloud data’, Automation in
Construction 127, 103703.
3. Kim, S., Yajima, Y., Park, J., Chen, J. & Cho, Y. K. (2022), ‘A hybrid semantic-geometric approach for clutter-resistant floorplan generation from building point clouds’. URL: https://doi.org/10.48550/arXiv.2305.15420
4. Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A. C., Lo, W.-Y. et al. (2023), Segment anything, in ‘2021 IEEE/CVF International Conference on Computer Vision (ICCV)’, IEEE.
5. Lee, T., Chan, K., Chan, H. & Kok, M. (2012), ‘Projections of extreme rainfall in hong kong in the 21st century’, Acta Meteorologica Sinica 25.
6. Maxime, Q., Stefan, B., Aymeric, H. & Laure, D. (2024), ‘Scan-to-bim: Unlocking current limitations through artificial intelligence’, in ‘2024 Proceedings of the 41st ISARC, Lille, France, ISBN 978-0-6458322-1-1, ISSN 2413-5844’, pp. 779-788. URL: https://www.iaarc.org/publications/fulltext/134_ISARC_2024_Paper_197.pdf
7. Perez-Perez, Y., Golparvar-Fard, M. & El-Rayes, K. (2021), ‘Scan2bim-net: Deep learning method for segmentation of point clouds for scan-to-bim’, Journal of Construction Engineering and Management 147(9). Publisher Copy- right: © 2021 American Society of
Civil Engineers.
8. Redmon, J., Divvala, S., Girshick, R. & Farhadi, A. (2016), You only look once: Unified, real-time object detection, in ‘2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)’, pp. 779–788.
9. Smith, J., Brown, K. et al. (2017), ‘Impacts of extreme weather on infrastructure and society: A review of recent research’, Journal of the Air Waste Management Association 67(4), 445–460. URL:https://www.tandfonline.com/doi/full/10.1080/10962247.2017.1401017
10. Xiong, X., Adan, A., Akinci, B. & Huber, D. (2013), ‘Automatic creation of semantically rich 3d building models from laser scanner data’, Automation in Construction 31, 325–337. URL: https://www.sciencedirect.com/science/article/pii/S0926580512001732
>> more<< less