Bidirectional Retrieval on SARVLM-1M (test)
30.94Mean RecallSARCLIP †
Evaluation Results
| Method | Links | |
|---|---|---|
| SARCLIP †Image Backbone=ViT-L-14, Pretrain Data=RS5M, Tune On=SARVLM-1M2025.10 | 30.94 | |
| SARCLIP ‡Image Backbone=ViT-L-14, Pretrain Data=R3+D10+S4, Tune On=SARVLM-1M2025.10 | 30.45 | |
| SARCLIP §Image Backbone=ViT-L-14, Pretrain Data=SkyScript, Tune On=SARVLM-1M2025.10 | 30.36 | |
| SARCLIPImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=SARVLM-1M2025.10 | 29.96 | |
| SARCLIP †Image Backbone=ViT-B-32, Pretrain Data=RS5M, Tune On=SARVLM-1M2025.10 | 27.66 | |
| SARCLIP ‡Image Backbone=ViT-B-32, Pretrain Data=R3+D10+S4, Tune On=SARVLM-1M2025.10 | 27.47 | |
| SARCLIPImage Backbone=ViT-B-32, Pretrain Data=LAION, Tune On=SARVLM-1M2025.10 | 27.15 | |
| SARCLIP §Image Backbone=ViT-B-32, Pretrain Data=SkyScript, Tune On=SARVLM-1M2025.10 | 26.89 | |
| SARCLIP ‡Image Backbone=ResNet-50, Pretrain Data=R3+D10+S4, Tune On=SARVLM-1M2025.10 | 24.79 | |
| SARCLIPImage Backbone=ResNet-50, Pretrain Data=LAION, Tune On=SARVLM-1M2025.10 | 24.75 | |
| SAR-TEXTImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=SARTEXT2025.10 | 1.52 | |
| SARCLIP-ISPRSImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=SARCAP2025.10 | 1.46 | |
| SARCLIP-ISPRSImage Backbone=ViT-B-32, Pretrain Data=–, Tune On=–2025.10 | 0.91 | |
| SkyCLIPImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=SkyScript2025.10 | 0.61 | |
| GeoRSCLIPImage Backbone=ViT-H-14, Pretrain Data=LAION, Tune On=RS5M2025.10 | 0.54 | |
| GeoRSCLIPImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=RS5M2025.10 | 0.52 | |
| SARCLIP-ISPRSImage Backbone=ResNet-50, Pretrain Data=LAION, Tune On=SARCAP2025.10 | 0.51 | |
| SkyCLIPImage Backbone=ViT-B-32, Pretrain Data=LAION, Tune On=SkyScript2025.10 | 0.5 | |
| OpenCLIPImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=–2025.10 | 0.49 | |
| RemoteCLIPImage Backbone=ViT-L-14, Pretrain Data=LAION, Tune On=R3+D10+S42025.10 | 0.46 | |
| GeoRSCLIPImage Backbone=ViT-B-32, Pretrain Data=LAION, Tune On=RS5M2025.10 | 0.43 | |
| RemoteCLIPImage Backbone=ResNet-50, Pretrain Data=LAION, Tune On=R3+D10+S42025.10 | 0.41 | |
| RemoteCLIPImage Backbone=ViT-B-32, Pretrain Data=LAION, Tune On=R3+D10+S42025.10 | 0.28 | |
| OpenCLIPImage Backbone=ViT-B-32, Pretrain Data=LAION, Tune On=–2025.10 | 0.27 | |
| OpenCLIPImage Backbone=ResNet-50, Pretrain Data=LAION, Tune On=–2025.10 | 0.21 |