EarthInfer
Agentic reasoning with dynamically generated modalities for Earth observation
Overview
EarthInfer is an agentic framework for answering complex questions about environmental risk and land analysis from satellite imagery. It combines TerraMind with autonomous Earth-observation data retrieval and spatiotemporal reasoning.
Rather than relying only on the original image, the system generates task-relevant geospatial modalities such as vegetation indices and land-use or land-cover maps during reasoning.
Publication
Presented at the WACV 2026 Workshop on Computer Vision for Earth Observation (CV4EO).