N
Number Six
Zero-Shot Multi-Agent LLM Pipeline for Disaster Damage Mapping from Satellite and Street-View Imagery
A disaster damage mapping workflow in which cooperating multimodal large language model agents enhance, interpret, and classify paired satellite and street-view images without task-specific training, then georeference the results into a web GIS map. Students learn how to decompose a remote sensing assessment task into specialised agent roles and how to evaluate zero-shot damage classification against ground-truth labels.