Exploring the Potential of Contrastive Language-Image Pre-training for Multi-Source Remote Sensing Data

arXiv:2609.03391v2 Announce Type: replace-cross Abstract: Contrastive language-image learning (CLIP) has become a key paradigm for remote sensing vision-language understanding. However, existing remote sensing contrastive learning methods are mostly built on R