Image Classification on RSI-CB128
49.4AccuracyT-SkyCLIP50
Evaluation Results
| Method | Links | |
|---|---|---|
| T-SkyCLIP50Setting=Transductive, Transductive Inference (TransCLIP)=true2025.06 | 49.4 | |
| SkyCLIPArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 48.5 | |
| SOTASetting=Transductive, Transductive Inference (TransCLIP)=false2025.06 | 47.6 | |
| SkyCLIPARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 43.9 | |
| T-GeoRSCLIPSetting=Transductive, Transductive Inference (TransCLIP)=true2025.06 | 42.8 | |
| CLIP LAION-RSArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 42.3 | |
| GeoRSCLIPArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 41.4 | |
| GeoRSCLIPARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 41.3 | |
| MetaCLIP-2.5BARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 40.7 | |
| CLIPBackbone=ViT-L-14, zero-shot=true2023.06 | 40.23 | |
| CLIPBackbone=ViT-L-14, zero-shot=true2023.06 | 40.23 | |
| CLIPArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 40.2 | |
| CLIP LAION-RSARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 39.8 | |
| SigLIP2Architecture=ViT-L/16, Setting=Zero-shot, Template=DST2026.06 | 39.8 | |
| MetaCLIP-2.5BArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 39.5 | |
| SkyCLIP50Setting=Transductive, Transductive Inference (TransCLIP)=false2025.06 | 39 | |
| SigLIP2ARCH=ViT-L/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 37.8 | |
| RemoteCLIPBackbone=ViT-L, zero-shot=true2025.12 | 37.23 | |
| RemoteCLIPBackbone=ViT-L-14, zero-shot=true2023.06 | 37.22 | |
| RemoteCLIPBackbone=ViT-L-14, zero-shot=true2023.06 | 37.22 | |
| GeoMELTBackbone=BEiT-3, zero-shot=true2025.12 | 36.13 | |
| MetaCLIP-400MArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 35.5 | |
| MetaCLIP-2.5BARCH=ViT-B/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 35.4 | |
| MetaCLIP-400MARCH=ViT-B/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 34.8 | |
| SigLIP2Architecture=ViT-B/16, Setting=Zero-shot, Template=DST2026.06 | 34.5 | |
| GeoRSCLIPARCH=ViT-B/32, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 34.1 | |
| RemoteCLIPArchitecture=ViT-L/14, Setting=Zero-shot, Template=DST2026.06 | 33.9 | |
| MetaCLIP-400MArchitecture=ViT-B/16, Setting=Zero-shot, Template=DST2026.06 | 33.6 | |
| T-CLIPSetting=Transductive, Transductive Inference (TransCLIP)=true2025.06 | 33.2 | |
| SigLIPARCH=ViT-L/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 33.2 | |
| SigLIP2ARCH=ViT-B/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 33.2 | |
| MetaCLIP-2.5BArchitecture=ViT-B/16, Setting=Zero-shot, Template=DST2026.06 | 33.2 | |
| CLIPARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 32.7 | |
| CLIPBackbone=ViT-L, zero-shot=true2025.12 | 32.69 | |
| RemoteCLIPARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 31.4 | |
| T-RemoteCLIPSetting=Transductive, Transductive Inference (TransCLIP)=true2025.06 | 31.1 | |
| CLIPArchitecture=ViT-B/16, Setting=Zero-shot, Template=DST2026.06 | 30.9 | |
| MetaCLIP-400MARCH=ViT-L/14, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 30.8 | |
| SigLIPARCH=ViT-B/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 30.3 | |
| SigLIPArchitecture=ViT-L/16, Setting=Zero-shot, Template=DST2026.06 | 30.3 | |
| GeoRSCLIPArchitecture=ViT-B/32, Setting=Zero-shot, Template=DST2026.06 | 29.9 | |
| GeoRSCLIPSetting=Transductive, Transductive Inference (TransCLIP)=false2025.06 | 29 | |
| CLIPBackbone=ViT-B-32, zero-shot=true2023.06 | 28.88 | |
| CLIPBackbone=ViT-B-32, zero-shot=true2023.06 | 28.88 | |
| CLIPARCH=ViT-B/16, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 28.6 | |
| CLIPSetting=Transductive, Transductive Inference (TransCLIP)=false2025.06 | 27.7 | |
| RemoteCLIPARCH=ViT-B/32, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 27.7 | |
| CLIPBackbone=ResNet-50, zero-shot=true2023.06 | 26.56 | |
| CLIPBackbone=ResNet-50, zero-shot=true2023.06 | 26.56 | |
| SigLIPArchitecture=ViT-B/16, Setting=Zero-shot, Template=DST2026.06 | 26.2 | |
| RemoteCLIPArchitecture=ViT-B/32, Setting=Zero-shot, Template=DST2026.06 | 26.2 | |
| RemoteCLIPSetting=Transductive, Transductive Inference (TransCLIP)=false2025.06 | 26 | |
| RemoteCLIPBackbone=ViT-B, zero-shot=true2025.12 | 25.88 | |
| RemoteCLIPBackbone=ViT-B-32, zero-shot=true2023.06 | 24.18 | |
| RemoteCLIPBackbone=ViT-B-32, zero-shot=true2023.06 | 24.18 | |
| RemoteCLIPARCH=RN-50, Zero-shot=true, Approach=Meta-prompting (LLaMA-3.1)2026.06 | 22 | |
| CLIPBackbone=ViT-B, zero-shot=true2025.12 | 20.09 | |
| RemoteCLIPArchitecture=RN-50, Setting=Zero-shot, Template=DST2026.06 | 16.3 | |
| RemoteCLIPBackbone=ResNet-50, zero-shot=true2023.06 | 13.95 | |
| RemoteCLIPBackbone=ResNet-50, zero-shot=true2023.06 | 13.95 |