Publications / 2024 Proceedings of the 41st ISARC, Lille, France

Semantic annotation of images from outdoor construction sites

Layan Farahat, Ehsan Rezazadeh Azar
Pages 745-752 (2024 Proceedings of the 41st ISARC, Lille, France, ISBN 978-0-6458322-1-1, ISSN 2413-5844)
Abstract:

Valuable information is embedded in construction images which can be used for different construction engineering and management purposes. The availability of low-cost cameras and robust artificial intelligence methods has increased the use of imaging technology in construction sites. However, these rich data sources are not often used to their full potential due to subjective documentation, leading to potentially overlooking valuable content. This study proposes an ensemble approach that utilizes deep learning techniques for object recognition, pixel-level segmentation, and text classification to annotate still images from outdoor construction scenes at medium (ongoing activities) and high (project type) levels. Experimental results demonstrate the potential of this approach by achieving a 70% overall recall rate.

Keywords: Scene understanding, construction images, construction management, deep learning, object detection, semantic segmentation, annotation