GAIR introduces a geo-aligned implicit representation module inside a multi-encoder contrastive SSL framework that produces location-aware embeddings and outperforms prior geo-foundation models on 22 geospatial datasets across 9 tasks.
Senclip: Enhancing zero-shot land-use mapping for sentinel-2 with ground-level prompting
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GAIR: Location-Aware Self-Supervised Contrastive Pre-Training with Geo-Aligned Implicit Representations
GAIR introduces a geo-aligned implicit representation module inside a multi-encoder contrastive SSL framework that produces location-aware embeddings and outperforms prior geo-foundation models on 22 geospatial datasets across 9 tasks.