V2V-AI
Geospatial AI for assessing residential conditions in Detroit.
V2V-AI explores how street-level imagery, geospatial data, computer vision, and vision-language models can support scalable assessment of residential conditions in Detroit.
Research focus
The project asks how AI-derived information about the built environment can complement existing urban data while remaining transparent, useful, and responsible. Current work includes the UrbanWORM Python package and vision-language modeling of residential property and blight conditions.
Partnership
The work is being developed with City of Detroit partners and is oriented toward practical urban analysis and decision-making rather than automated classification as an end in itself.
Methods
Street-level imagery · GeoAI · computer vision · vision-language models · spatial data integration · responsible AI