New preprint examines vision-language models for urban blight assessment

Detroit-focused work tests whether vision-language models can extract useful information on residential property conditions.

A new preprint, “Can Urban Blight Be Accessed with Vision-language Models: A Case Study in Detroit,” examines the use of vision-language models to infer residential property conditions from street-level imagery.

The work contributes to a broader collaboration on scalable, transparent urban analytics and decision support in Detroit.