Rueda-Plata, Diego ; Gonzales, Daniela ; Acevedo, Ana B; Duque, Juan C; Ramos-Pollán, Raúl. Use of deep learning models in street-level images to classify one-story unreinforced masonry buildings based on roof diaphragms Artículo de revista Building and Environment, 189 (107517), pp. 1-10, 2020, ISSN: 0360-1323. Resumen | Enlaces | BibTeX | Etiquetas: Redes neuronales, Sostenibilidad de ciudades @article{Rueda2020,
title = {Use of deep learning models in street-level images to classify one-story unreinforced masonry buildings based on roof diaphragms},
author = {Rueda-Plata, Diego and Gonzales, Daniela and Acevedo, Ana B. and Duque, Juan C. and Ramos-Pollán, Raúl.},
url = {https://doi.org/10.1016/j.buildenv.2020.107517},
doi = {10.1016/j.buildenv.2020.107517},
issn = {0360-1323},
year = {2020},
date = {2020-12-21},
journal = {Building and Environment},
volume = {189},
number = {107517},
pages = {1-10},
abstract = {In this paper, we explore the potential of convolutional neural networks to classify street-level imagery of one-story unreinforced masonry buildings (MURs) according to the flexibility of the roof diaphragm (rigid or flexible). This information is critical for vulnerability studies, disaster risk assessments, disaster management strategies, etc., and is of great relevance in cities where unreinforced masonry is the most common building typology or where the majority of the population resides in such buildings. Our contribution could be useful for local governments of cities in developing countries seeking to significantly reduce the number of deaths caused by disasters. Our research results indicate that VGG19 is the convolutional neural network architecture with the best performance, with an accuracy of 0.80, a precision of 0.88, and a recall of 0.84. The results are encouraging and could be used to reduce the amount of resources (both human and economic) for the development of detailed exposure models for unreinforced masonry buildings.},
keywords = {Redes neuronales, Sostenibilidad de ciudades},
pubstate = {published},
tppubtype = {article}
}
In this paper, we explore the potential of convolutional neural networks to classify street-level imagery of one-story unreinforced masonry buildings (MURs) according to the flexibility of the roof diaphragm (rigid or flexible). This information is critical for vulnerability studies, disaster risk assessments, disaster management strategies, etc., and is of great relevance in cities where unreinforced masonry is the most common building typology or where the majority of the population resides in such buildings. Our contribution could be useful for local governments of cities in developing countries seeking to significantly reduce the number of deaths caused by disasters. Our research results indicate that VGG19 is the convolutional neural network architecture with the best performance, with an accuracy of 0.80, a precision of 0.88, and a recall of 0.84. The results are encouraging and could be used to reduce the amount of resources (both human and economic) for the development of detailed exposure models for unreinforced masonry buildings. |
Gonzalez, Daniela ; Rueda-Plata, Diego ; Acevedo, Ana B; Duque, Juan C; Ramos-Pollán, Raúl ; Betancourt, Alejandro ; García, Sebastian Automatic detection of building typology using deep learning methods on street level image Artículo de revista Building and Environment, 177 (106805), pp. 1-12, 2020, ISSN: 0360-1323. Resumen | Enlaces | BibTeX | Etiquetas: Redes neuronales, Sostenibilidad de ciudades @article{gonzalez2020automatic,
title = {Automatic detection of building typology using deep learning methods on street level image},
author = {Gonzalez, Daniela and Rueda-Plata, Diego and Acevedo, Ana B. and Duque, Juan C. and Ramos-Pollán, Raúl and Betancourt, Alejandro and García, Sebastian},
url = {https://doi.org/10.1016/j.buildenv.2020.106805},
doi = {10.1016/j.buildenv.2020.106805},
issn = {0360-1323},
year = {2020},
date = {2020-06-15},
journal = {Building and Environment},
volume = {177},
number = {106805},
pages = {1-12},
abstract = {An exposure model is a key component for assessing potential human and economic losses from natural disasters. An exposure model consists of a spatially disaggregated description of the infrastructure and population of a region under study. Depending on the size of the settlement area, developing such models can be a costly and time-consuming task. In this paper we use a manually annotated dataset consisting of approximately 10,000 photos acquired at street level in the urban area of Medellín to explore the potential for using a convolutional neural network (CNN) to automatically detect building materials and types of lateral-load resisting systems, which are attributes that define a building's structural typology (which is a key issue in exposure models for seismic risk assessment). The results of the developed model achieved a precision of 93% and a recall of 95% when identifying nonductile buildings, which are the buildings most likely to be damaged in an earthquake. Identifying fine-grained material typology is more difficult because many visual clues are physically hidden, but our model matches expert level performances, achieving a recall of 85% and accuracy scores ranging from 60% to 82% on the three most common building typologies, which account for 91% of the total building population in Medellín. Overall, this study shows that a CNN can make a substantial contribution to developing cost-effective exposure models.},
keywords = {Redes neuronales, Sostenibilidad de ciudades},
pubstate = {published},
tppubtype = {article}
}
An exposure model is a key component for assessing potential human and economic losses from natural disasters. An exposure model consists of a spatially disaggregated description of the infrastructure and population of a region under study. Depending on the size of the settlement area, developing such models can be a costly and time-consuming task. In this paper we use a manually annotated dataset consisting of approximately 10,000 photos acquired at street level in the urban area of Medellín to explore the potential for using a convolutional neural network (CNN) to automatically detect building materials and types of lateral-load resisting systems, which are attributes that define a building's structural typology (which is a key issue in exposure models for seismic risk assessment). The results of the developed model achieved a precision of 93% and a recall of 95% when identifying nonductile buildings, which are the buildings most likely to be damaged in an earthquake. Identifying fine-grained material typology is more difficult because many visual clues are physically hidden, but our model matches expert level performances, achieving a recall of 85% and accuracy scores ranging from 60% to 82% on the three most common building typologies, which account for 91% of the total building population in Medellín. Overall, this study shows that a CNN can make a substantial contribution to developing cost-effective exposure models. |