GeoMeld provides a large-scale aligned multimodal remote sensing dataset with verified semantic captions and a joint pretraining method that improves downstream transfer and cross-sensor robustness in foundation models.
Rs5m and georsclip: A large-scale vision-language dataset and a large vision-language model for remote sensing.IEEE Transactions on Geoscience and Remote Sensing, 62:1–23,
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
-
GeoMeld: Toward Semantically Grounded Foundation Models for Remote Sensing
GeoMeld provides a large-scale aligned multimodal remote sensing dataset with verified semantic captions and a joint pretraining method that improves downstream transfer and cross-sensor robustness in foundation models.