Streetview Imaging for Automated Assessments of Urban Infrastructure and Services

Abstract

Many forms of ambient data in cities are starting to become available that allow tracking of short-term urban operations, such as traffic management, trash collections, inspections or non-emergency maintenance requests. However, arguably the greatest promise of urban analytics is to set up measurable objectives and track progress towards systemic development goals connected to human development and sustainability over the longer term. The challenge for such an approach is the connection between new technological capabilities, such as sensing and machine learning, and local knowledge and operations of residents and city governments. Here we describe an emerging project for the long-term monitoring of sustainable development in fast growing towns in the Galapagos through the convergence of these methods.We demonstrate how collaborative mapping and the capture of 360 degree street views can produce a general basis for a broad set of quantitative analytics, when such actions are coupled to mapping and deep learning characterization of urban environments. We map and assess the precision of urban assets via automatic object classification and characterize their abundance and spatial heterogeneity. We also discuss how these methods, as they continue to improve, can provide the means to perform an ambient census of urban assets (buildings, vehicles, services) and environmental conditions.