What is it about?

Gentrification is a complicated process that involves changes to neighborhoods, making them more upscale. One clear sign of gentrification is visible improvements to a street. Thanks to recent technology developments in Artificial Intelligence, we can now use computer vision techniques to look at every house through time in an entire city and then create heat-maps to show where gentrification is happening on hitherto unprecedented geographic scales. In this study, we used a special type of artificial intelligence (AI) called a Siamese Convolutional Neural Network (SCNN) to detect visual changes related to gentrification in Google Street View images over time. We analyzed images from over 86,000 individual properties in Ottawa, Canada, between 2007 and 2016. Our AI system was highly accurate (95.6%) in identifying these changes. We then created maps that showed which areas had the most visible improvements to properties. These maps matched up well with building permit data from the City of Ottawa between 2011 and 2016. Our maps confirmed the areas known to be undergoing gentrification and even discovered new areas experiencing it. Unlike previous studies, our approach focused on individual properties and their visual improvements over time. The heat-maps illustrated regions within the city that had a high concentration of these visual gentrification-like changes.

Featured Image

Why is it important?

The study is important for several reasons: Innovative approach: By using advanced AI techniques like Siamese Convolutional Neural Networks (SCNN) and Kernel Density Estimation (KDE), the study offers a novel and efficient way to detect and map gentrification. This approach is more detailed and accurate than traditional methods, as it focuses on individual properties and their visual improvements over time. Increased awareness: The study helps increase awareness about the ongoing process of gentrification in urban areas. By providing clear, visual evidence of these changes, it enables policymakers, urban planners, and community members to better understand the extent and patterns of gentrification. Informed decision-making: The results of the study can support policymakers and urban planners in making informed decisions when addressing gentrification issues. By identifying areas undergoing gentrification, they can develop targeted strategies and interventions to manage the process and its impact on local communities. Revealing unknown areas: The study not only confirms areas known to be undergoing gentrification but also uncovers previously unknown areas experiencing it. This information is valuable for researchers and city officials to better understand the dynamics of urban change and the factors driving gentrification. Encouraging further research: The study's innovative approach and promising results can inspire other researchers to adopt similar methods or explore new ways of using AI and computer vision to study urban phenomena, such as housing affordability, neighborhood change, and urban development.


This work can be seen as a significant contribution to the field of urban studies and an innovative application of AI and computer vision techniques. Perhaps this will inspire further applications of these methods in studying other urban phenomena and social issues. Community activists and local residents might view the study's findings as crucial information to raise awareness about gentrification and its consequences on local communities. The work can serve as a basis for advocacy, promoting equitable urban development and addressing issues such as displacement and housing affordability. And finally, this study demonstrates the potential of AI to address real-world problems. It can be seen as a successful case of technology being applied to social issues, showcasing the benefits and capabilities of AI in various domains.

University of Ottawa

Read the Original

This page is a summary of: Deep mapping gentrification in a large Canadian city using deep learning and Google Street View, PLoS ONE, March 2019, PLOS, DOI: 10.1371/journal.pone.0212814.
You can read the full text:

Open access logo


The following have contributed to this page