AI Analyzes Google Street View Images to Reveal Hidden Building Details, Aiding Urban Planning and Resource Management

Researchers at the University of Toronto have pioneered a method to extract deeper insights about buildings using artificial intelligence (AI) and Google Street View images. The technique, developed by Associate Professor Shoshanna Saxe and Senior AI Researcher Alex Olson, aims to predict key attributes like the age and floor area of buildings, providing valuable data for urban planning and sustainability research. The study, published in the Journal of Industrial Ecology, leverages the widespread availability of Google Street View images to offer a cost-effective solution for generating large-scale building data. Traditionally, collecting detailed information about buildings, such as their age and internal dimensions, requires significant resources and expertise, often costing millions of dollars. However, the U of T team’s AI model can produce similar data for a fraction of the cost. The researchers trained their AI algorithm to estimate building attributes by analyzing exterior images. They achieved 70% accuracy in predicting a building's age and 80% accuracy in predicting its floor area. These predictions are not just surface-level deductions; they provide a foundation for understanding the internal uses and resource consumption of buildings. For example, knowing a building’s age helps in estimating the types of materials used in its construction and the associated embodied greenhouse gases, which are emissions generated during the production and transportation of construction materials. Saxe emphasizes that their approach goes beyond what can be inferred from maps or building plans alone. “You need to see structures,” she explains. “We’re predicting the internal square footage, which is harder to estimate just by looking at the outside. Renovations and varying materials on different parts of the building can make age determination challenging.” By integrating this AI-driven data with existing urban datasets, urban planners can gain a more comprehensive view of the city’s infrastructure. This information can help identify areas with underutilized resources or infrastructure, inform future planning decisions, and optimize resource allocation. According to Olson, “While it doesn’t model the future, it does provide an accurate description of the current situation, allowing us to use the data for planning resource uses and future projects.” The method has broad applications beyond carbon assessment. Saxe mentions that researchers have shown interest in using this data for water usage studies and resilience assessments. The ability to predict the internal characteristics of buildings can also enhance energy efficiency analyses and support initiatives aimed at reducing urban environmental impacts. One of the key advantages of this approach is its scalability. The low cost and ease of data collection enable researchers and urban planners to analyze entire neighborhoods or even cities. This scalability is particularly crucial for large urban areas where comprehensive data collection would otherwise be prohibitively expensive. The implications of this research are significant for both academic and practical purposes. For instance, it can aid in the development of more sustainable urban environments by providing data-driven insights into building performance and material use. It also supports smarter resource management, ensuring that infrastructure projects align with the actual needs of communities. Industry insiders have praised the innovation for its potential to revolutionize urban planning and sustainability research. The University of Toronto's Center for the Sustainable Built Environment, where Saxe leads the research, is known for its cutting-edge work in sustainable infrastructure. Similarly, the Center for Analytics & Artificial Intelligence Engineering, where Olson is based, focuses on developing advanced AI solutions for real-world problems. In summary, the U of T team's AI-driven method for extracting detailed building information from Google Street View images is a groundbreaking tool that can enhance urban planning, resource management, and sustainability efforts. Its cost-effectiveness and scalability make it particularly valuable for large urban areas, and its versatile applications promise to open new avenues for research and practical implementation in the field of urban ecology.