SatBLIP fine-tunes a satellite-adapted BLIP model on GPT-4o-generated captions to predict county-level SVI from satellite tiles and uses SHAP to highlight key features like roof condition and vegetation.
2024.GeoLLM-Engine: A Realistic Environment for Building Geospatial Copilots
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SatBLIP: Context Understanding and Feature Identification from Satellite Imagery with Vision-Language Learning
SatBLIP fine-tunes a satellite-adapted BLIP model on GPT-4o-generated captions to predict county-level SVI from satellite tiles and uses SHAP to highlight key features like roof condition and vegetation.