We analyzed Philly street scenes and identified signs of gentrification using machine learning trained on longtime residents’ observations
Summary
Ph.D. candidates from Drexel and Temple universities developed a new method to map gentrification in Philadelphia by combining long-term resident accounts, Google Street View imagery, and machine learning. The research team conducted focus groups in three rapidly gentrifying neighborhoods, asking residents to identify visual cues of gentrification, such as specific building designs, materials, and colors. These qualitative descriptions were then corroborated with historical Google Street View images from 2009-13 and 2017-21, translating generalized terms like "modern" into architectural specifics like "presence of bump-out windows." A deep mapping model, utilizing neural network algorithms, was trained to distinguish "gentrified" from "not-gentrified" scenes with approximately 84% accuracy, demonstrating that resident observations can be reliably translated into machine learning signals. The study highlights how AI can help identify and predict physical environmental changes associated with gentrification, offering a tool for neighborhood stakeholders and researchers.
Key takeaway
For urban planners and community organizers tracking neighborhood change, this research offers a robust method to quantify gentrification's visual impact. You can leverage AI-driven deep mapping models, informed by local resident input, to identify specific architectural changes and predict future development trajectories. This approach provides objective data to support community concerns and inform policy decisions, moving beyond anecdotal evidence to data-backed insights on environmental and social shifts.
Key insights
Resident perceptions of gentrification can be effectively translated into machine learning signals for urban analysis.
Principles
- Local resident input is crucial for defining gentrification's visual cues.
- Deep mapping models can accurately categorize urban scenes based on learned patterns.
Method
The method combines resident focus group input, historical Google Street View imagery analysis, and deep mapping models to identify and categorize visual signs of gentrification with high accuracy.
In practice
- Use deep mapping models to identify gentrification hot spots.
- Corroborate resident observations with historical imagery.
- Apply XAI to understand model prediction rationale.
Topics
- Gentrification Mapping
- Machine Learning
- Google Street View
- Resident Perceptions
- Deep Mapping Models
Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Policy Maker
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial intelligence (AI) – The Conversation.