Download PDFOpen PDF in browserAI-based Object Detection and Material Classification for Scan-to-BIM – A Deep Learning Framework10 pages•Published: July 23, 2025AbstractBuilding Information Modeling (BIM) is a pivotal technology in the Architecture, Engineering, and Construction (AEC) industry. It enables the creation and management of digital representations of physical and functional characteristics of buildings. The integration of deep learning technology can significantly enhance BIM’s capabilities. This is particularly true for object detection and material classification. This paper explores the application of deep learning techniques, including Convolutional Neural Networks (CNNs), for the automated detection and classification of building elements and materials. We detail the workflows in training and deploying these models, encompassing data collection, preprocessing, model training, and validation. Furthermore, we discuss the integration of these models into BIM frameworks. Emphasizing the benefits such as improved accuracy, efficiency, and cost-effectiveness. There are challenges like large data requirements and computational demands. Still, the potential for deep learning to transform BIM processes is immense. This article suggests a framework as part of ongoing research to address current limitations and advance the automation of Scan-to-BIM processes. It introduces sophisticated deep-learning models for object detection and material identification within the construction environment. The proposed framework will be implemented, and future research articles will present the final results.Keyphrases: aec industry, deep learning, material classification, object detection, scan to bim In: Wesley Collins, Anthony J. Perrenoud and John Posillico (editors). Proceedings of Associated Schools of Construction 61st Annual International Conference, vol 6, pages 430-439.
|
