We analyzed Philly street scenes and identified signs of gentrification using machine learning trained on longtime residents’ observations

New high-rise, modern apartment buildings that starkly contrast with traditional row homes are key signs of gentrification in Philadelphia, often described by longtime residents as “out of place” and lacking architectural style.
Researchers at Drexel and Temple universities developed a method to map gentrification using residents’ descriptions and analysis of Google Street View images and machine learning techniques.
Focus groups revealed that longtime residents could pinpoint visual cues associated with gentrification, such as changes in building design, materials, colors, and landscaping, corroborated by historical Street View data.
They used a deep mapping AI model to determine differences in building scenery from pre- and post-gentrification time periods, achieving about 84% accuracy in identifying gentrified areas.
Researchers emphasize the importance of transparency in AI models that predict gentrification, utilizing techniques like explainable AI (XAI) to help understand the model’s decision-making process and ensure its insights are relevant and accurate.