Automatic identification of residential building features using machine learning techniques

Seismic risk assessment represents a major challenge in countries with considerable seismic hazard and vulnerable built heritage, such as Italy. Especially for large-scale analyses, vulnerability evaluation may result in very time-consuming and expensive investigations. In order to tackle this problem, satellite images and street view pictures can be used to automatically collect data about buildings through Convolutional Neural Networks (CNNs). In this way, meaningful parameters that influence seismic vulnerability can be acquired remotely from images, with a significant reduction in time and costs. This can also lead to a better vulnerability assessment, thus to more precise seismic risk estimates.