• Robust building identification from street views using deep convolutional neural networks

Roussel, Robin ORCID: https://orcid.org/0000-0001-8875-3688, Jacoby, Sam ORCID: https://orcid.org/0000-0002-9133-5177 and Asadipour, Ali ORCID: https://orcid.org/0000-0003-0159-3090, 2024, Journal Article, Robust building identification from street views using deep convolutional neural networks Buildings, 14 (3). pp. 1-26. ISSN 2075-5309

Abstract or Description:

Street view imagery (SVI) is a rich source of information for architectural and urban analysis using computer vision techniques, but its integration with other building-level data sources requires an additional step of visual building identification. This step is particularly challenging in architecturally homogeneous, dense residential streets featuring narrow buildings, due to a combination of SVI geolocation errors and occlusions that significantly increase the risk of confusing a building with its neighboring buildings. This paper introduces a robust deep learning-based method to identify buildings across multiple street views taken at different angles and times, using global optimization to correct the position and orientation of street view panoramas relative to their surrounding building footprints. Evaluating the method on a dataset of 2000 street views shows that its identification accuracy (88%) outperforms previous deep learning-based methods (79%), while methods solely relying on geometric parameters correctly show the intended building less than 50% of the time. These results indicate that previous identification methods lack robustness to panorama pose errors when buildings are narrow, densely packed, and subject to occlusions, while collecting multiple views per building can be leveraged to increase the robustness of visual identification by ensuring that building views are consistent.

Official URL: https://www.mdpi.com/2075-5309/14/3/578
Subjects: Other > Mathematical and Computer Sciences > G400 Computer Science
Other > Mathematical and Computer Sciences > G700 Artificial Intelligence > G740 Computer Vision
Architecture > K400 Planning (Urban > K440 Urban studies
School or Centre: Research Centres > Computer Science Research Centre
School of Architecture
Funders: Prosit Philosophiae Foundation
Identification Number or DOI: https://doi.org/10.3390/buildings14030578
Uncontrolled Keywords: building identification; street view imagery; CNN
Date Deposited: 22 Feb 2024 10:41
Last Modified: 22 Feb 2024 10:41
URI: https://researchonline.rca.ac.uk/id/eprint/5736
Edit Item (login required) Edit Item (login required)