ResearchBenchmarksImage-to-Text Retrieval on GeoSound Bing imagery, scale=1Follow90.8R@10%Sat2Sound31.72847.06462.477.736May 19, 2025Evaluation ResultsMethodMethodLinksR@10%MRSat2Soundmetadata=w metametadata=w meta2025.0590.8160Sat2Soundmetadata=w/o metametadata=w/o meta2025.0588.1183CLIPModel=CLIPModel=CLIP2025.0552.81,420SigLIP2Model=SigLIP2Model=SigLIP22025.0544.92,641SigLIPModel=SigLIPModel=SigLIP2025.05343,307