Szczegóły publikacji
Opis bibliograficzny
Framework for standardizing BIM-based dataset creation for deep learning / K. TOMASZKIEWICZ, T. OWERKO // W: Moving toward smart, resilient and sustainable bridges : proceedings of the 13th International Conference on Bridge Maintenance, Safety and Management : IABMAS 2026 : Orlando, Florida, USA, 6–10 July 2026 / ed. by F. Necati Catbas, Dan M. Frangopol, Hae-Bum Yun. — Boca Raton ; London ; New York : CRC Press ; Taylor & Francis, 2026. — Dod. ISBN: 9781041311737. — ISBN: 9781041313687; e-ISBN: 9781003778677. — S. 2102–2108. — Bibliogr. s. 2107–2108, Abstr.
Autorzy (2)
Dane bibliometryczne
| ID BaDAP | 170068 |
|---|---|
| Data dodania do BaDAP | 2026-10-27 |
| Tekst źródłowy | URL |
| DOI | 10.1201/9781003778677-255 |
| Rok publikacji | 2026 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Creative Commons | |
| Wydawcy | CRC Press / Balkema, Taylor & Francis Group |
Abstract
The use of BIM information models in the maintenance phase is growing in popularity. At the same time, condition assessment processes (including damage detection) conducted during the maintenance phase are becoming increasingly supported by artificial intelligence-based solutions. However, no existing approaches enable the integration of information from technical condition assessment processes contained in BIM models with damage detection results obtained using deep learning algorithms. At the same time, the availability of properly labeled data that can be used as training and testing data for deep machine learning algorithms is very limited. This paper presents a framework for standardizing data stored in AIM (Asset Information Model) models and databases containing photographs from bridge inspections. The proposed solution is based on open standards from the construction industry and data science formats, and it complies with the requirements specified in the guidelines for conducting inspections and assessing technical conditions. It allows for the scalable and standardized creation of training and test datasets for deep learning-based solutions. In this way, it extends the applicability of deep neural networks not only for damage detection but also for tasks such as predicting the numerical rate of an element’s technical condition or determining the need for expertise. As a result, it is possible to develop solutions that support inspectors in their daily work and to store complete technical condition data in BIM models, along with photo databases, including annotations of element conditions and any observed damage.
