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

Ph.D. candidates at Drexel and Temple universities developed a machine learning model trained on longtime residents' descriptions of gentrification to identify visual signs of neighborhood change in Philadelphia street scenes. The researchers held focus groups in three rapidly gentrifying neighborhoods—one in Northeast Philadelphia and two in the River Wards—where residents described architectural markers like "high-rise, modern apartment buildings" and "gray, industrial looking" construction that contrasted with historic row homes. Using Google Street View images from 2009-13 and 2017-21, the team trained an AI model that achieved 84% accuracy in distinguishing gentrified from non-gentrified images by detecting pixel-level differences. The approach allows neighborhood groups and researchers to map development hotspots and quantify physical changes that residents observe, giving communities data to legitimize their concerns to media and city officials. The researchers say the model can help predict future gentrification trajectories and study connections between new construction and environmental conditions like air quality.