VLMs Map Urban Flooding, Identify 100k At-Risk New Yorkers Missed by Current Methods

2plus2make5 · x · 2026-08-28

A new paper in Nature Communications demonstrates using vision-language models (VLMs) to detect floods in large-scale street scene datasets. In New York City, the method identified flooded neighborhoods home to 100,000 people that current monitoring methods missed, highlighting VLMs' potential in environmental monitoring and urban disaster response.

Related event: VLM Maps Urban Flood Risk from Street View Imagery(2 posts)→

Original post →

More from Research

Research channel →