AI4DRR
This exploratory project proposes to develop a prototype platform that relies on artificial intelligence to automatically create exposure models suitable for risk assessment. The platform will rely on OSM to identify the footprint of each asset. Then, for each asset, the Google Street View API will be used to collect images from different angles, and a Deep Learning algorithm will be employed to estimate the most likely structural attributes. To train this algorithm, we will develop a database with photography of existing buildings with known structural attributes. Finally, the building footprint from OSM, the estimated structural parameters and building height from the satellite imagery will be combined using an Artificial Neural Network to create the digital representation of each asset. This prototype platform will be tested a parish in Lisbon, where ground proof data is available to evaluate the accuracy of the platform. Such results could enable the classification of the built environment for large regions, thus enabling risk assessment studies for a wide variety of hazards. This project has the support from the Humanitarian OpenStreetMap Team and the SimCenter AI group of California.
The project runs from 19-11-2021 to 17-11-2023 with a budget of 48.934,80 euros.
The following documents and deliverables are available:
Building's facade dataset: github.com/vsilva028/ML , https://zenodo.org/records/7625940
Building classification platform
First year report
Mobile application report
Jorge Lopes, MSc Dissertation report
Presentations:
UFP, Research seminar, 28th February 2023, presentation (video)
ICASP14 (Dublin), 9-13th July 2023, presentation
EPIA 2023 (Azores), 5-8th September 2023, presentation
SIBIU 2023 (Sibiu, Romania, Lucian Blaga University), 5-6th October 2023, presentation
UFP, Semana Ciência & Tecnologia, Ciência Viva, 23rd November 2023, presentation
UFP, Seminar “Improving disaster health and climate resilience at the expanded environmental crisis in Nepal and Bangladesh”, session “Best Practices and Innovation in Humanitarian Medicine”, 17 May 2024, presentation
Earthquake Spectra Showcase: May 2024 Issue, 13th June 2024, presentation
Journals:
Vítor Silva, Jorge Lopes, Feliz Gouveia, Gonçalo Lopes, "Assessment of Earthquake Risk Using Machine Learning for the Identification of Building Classes", Earthquake Spectra (in preparation).
Feliz Gouveia, Vítor Silva, Jorge Lopes, Rui Moreira, José Torres, Maria Guerreiro, “Automated Identification of Building Features with Deep learning for Risk Analysis”, SN Applied Sciences, Springer Nature, ISSN 2523-3971. (submitted)
Vitor Silva, Romain Sousa, Feliz Gouveia, Jorge Lopes, Maria João Guerreiro, “A Building Imagery Database for the Calibration of Machine Learning Algorithms”, Earthquake Spectra. February 8th, 2024. https://doi.org/10.1177/87552930241229103
Publications and communications:
Vitor Silva, Jorge Lopes, Feliz Gouveia, Romain Sousa, “Exposure Modelling through Machine Learning for Multi-Hazard Risk Assessment”, 14th International Conference on Applications of Statistics and Probability in Civil Engineering, ICASP14, 9-13 de Julho de 2023, Dublin, Irlanda (pdf). http://www.tara.tcd.ie/bitstream/handle/2262/103654/submission_553.pdf?sequence=1
Jorge Lopes, Feliz Gouveia, Vítor Silva, Rui Moreira, José Torres, Maria Guerreiro, Luís Paulo Reis, “Using Deep Learning for Building Stock Classification in Seismic Risk Analysis”, Encontro Português de Inteligência Artificial, EPIA 2023, 5-8 de Setembro de 2023, Ilha do Faial, Açores, Portugal (pdf). https://doi.org/10.1007/978-3-031-49011-8_41
Jorge Lopes, Feliz Gouveia, Vítor Silva, Rui Moreira, José Torres, Maria Guerreiro, Luís Paulo Reis, Using Deep Learning for Building Stock Classification in Seismic Risk Analysis, Sibiu Innovation Days (SID 2023), 6 de Outubro de 2023, Sibiu, Roménia. https://events.ulbsibiu.ro/innovationdays/presentations/
Vítor Silva, A Promessa das Novas Tecnologias na Avaliação do Risco Sísmico, 6ª Jornadas Portuguesas de Engenharia de Estruturas, Encontro Nacional de Betão Estrutural 2022, 12º Congresso Nacional de Sismologia e Engenharia Sísmica (SÍSMICA 2022), Lisboa, LNEC, 9 a 11 de Novembro de 2022. http://jpee2022.lnec.pt/palestras.html
Members:
Feliz Gouveia
Vítor Silva
Rui Moreira
José Torres
Luís Paulo Reis (LIACC / UP)
Jorge Lopes (MSc student)
Past members:
Muhammad Umar (MSc student)
Collaborators:
Romain Sousa
Maria João Guerreiro
Ivo Pereira
Nuno Moutinho (BSc student)
Daniel Moreira (BSc student)
Documents (Internal use)