Text Retrieval on RSICD (test)
22.14R@1GeoRSCLIP-FT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GeoRSCLIP-FTParadigm=GVLM + FT, Tuned on=RS5M + RSICD2023.06 | 22.14 | 40.53 | 51.78 | |
| GeoRSCLIP-FTParadigm=GVLM + FT, Tuned on=RS5M + RET-22023.06 | 21.13 | 41.72 | 55.63 | |
| RemoteCLIPParadigm=GVLM + FT, Tuned on=RET-3 + DET-10 + SEG-42023.06 | 17.02 | 37.97 | 51.51 | |
| MTGFEParadigm=Supervised, Tuned on=RSICD2023.06 | 15.28 | 37.05 | 51.6 | |
| OursBackbone=CLIP(ViT-B-32), Params=0.16M2023.08 | 14.13 | 31.51 | 44.78 | |
| PE-RSITRParadigm=GVLM + FT, Tuned on=RSICD2023.06 | 14.13 | 31.51 | 44.78 | |
| Full fine-tuningBackbone=CLIP(ViT-B-32), Params=151M2023.08 | 13.54 | 30.83 | 43.46 | |
| UniAdapterBackbone=CLIP(ViT-B-32), Params=0.55M2023.08 | 12.65 | 30.81 | 42.74 | |
| AdaptFormerBackbone=CLIP(ViT-B-32), Params=0.17M2023.08 | 12.46 | 28.49 | 41.86 | |
| GeoRSCLIPParadigm=GVLM + FT, Tuned on=RS5M2023.06 | 11.53 | 28.55 | 39.16 | |
| Cross-Modal AdapterBackbone=CLIP(ViT-B-32), Params=0.16M2023.08 | 11.18 | 27.31 | 40.62 | |
| Single LanguageBackbone=CLIP(ViT-B-32), Params=151M2023.08 | 10.7 | 29.64 | 41.53 | |
| MultilanguageParadigm=Supervised, Tuned on=RSICD2023.06 | 10.7 | 29.64 | 41.53 | |
| GeoRSCLIPParadigm=GVLM + Pfeiffer, Tuned on=RS5M2023.06 | 10.61 | 24.79 | 36.51 | |
| FAAMIParadigm=Supervised, Tuned on=RSICD2023.06 | 10.44 | 22.66 | 30.89 | |
| PIRParadigm=Supervised, Tuned on=RSICD2023.06 | 9.88 | 27.26 | 39.16 | |
| GeoRSCLIPParadigm=GVLM + LoRA, Tuned on=RS5M2023.06 | 9.52 | 21.13 | 31.75 | |
| GeoRSCLIPParadigm=GVLM + Prefixtuning, Tuned on=RS5M2023.06 | 9.06 | 20.04 | 29.73 | |
| GeoRSCLIPParadigm=GVLM + UniPELT, Tuned on=RS5M2023.06 | 8.97 | 23.79 | 35.04 | |
| AdapterBackbone=CLIP(ViT-B-32), Params=0.17M2023.08 | 8.73 | 24.73 | 37.81 | |
| CABIRBackbone=ResNet152, BERT+biGRU, Params=Not specified2023.08 | 8.59 | 16.27 | 24.13 | |
| Linear probeBackbone=CLIP(ViT-B-32), Params=1.05M2023.08 | 8.46 | 24.41 | 37.72 | |
| HVSAParadigm=Supervised, Tuned on=RSICD2023.06 | 7.47 | 20.62 | 32.11 | |
| SWANParadigm=Supervised, Tuned on=RSICD2023.06 | 7.41 | 20.13 | 30.86 | |
| VPTBackbone=CLIP(ViT-B-32), Params=0.46M2023.08 | 7.23 | 19.4 | 30.77 | |
| CLIP-AdapterBackbone=CLIP(ViT-B-32), Params=0.52M2023.08 | 7.11 | 19.48 | 31.01 | |
| Zero-shot CLIPBackbone=CLIP(ViT-B-32), Params=0.00M2023.08 | 6.77 | 15.37 | 23.15 | |
| MCRNBackbone=ResNet18, biGRU, Params=52.35M2023.08 | 6.59 | 19.4 | 30.28 | |
| GaLR with MRBackbone=ResNet18, biGRU, Params=46.89M2023.08 | 6.59 | 19.85 | 31.04 | |
| GaLRParadigm=Supervised, Tuned on=RSICD2023.06 | 6.59 | 19.85 | 31.04 | |
| ConvpassBackbone=CLIP(ViT-B-32), Params=0.14M2023.08 | 6.54 | 19.67 | 32.78 | |
| GaLR w/o MRBackbone=ResNet18, biGRU, Params=46.89M2023.08 | 6.5 | 18.91 | 29.7 | |
| VL-PromptBackbone=CLIP(ViT-B-32), Params=0.47M2023.08 | 6.47 | 17.05 | 28.68 | |
| CoOpBackbone=CLIP(ViT-B-32), Params=0.10M2023.08 | 6.32 | 15.89 | 27.6 | |
| SCAN i-tBackbone=Faster R-CNN, biGRU, Params=13.68M2023.08 | 5.85 | 12.89 | 19.84 | |
| KCRParadigm=Supervised, Tuned on=RSICD2023.06 | 5.84 | 22.31 | 36.12 | |
| AMFMN-fusionBackbone=ResNet18, biGRU, Params=35.94M2023.08 | 5.39 | 15.08 | 23.4 | |
| AFMFNParadigm=Supervised, Tuned on=RSICD2023.06 | 5.39 | 15.08 | 23.4 | |
| CLIP-BaselineParadigm=GVLM, Tuned on=None2023.06 | 5.31 | 14.18 | 23.7 | |
| AMFMN-simBackbone=ResNet18, biGRU, Params=35.94M2023.08 | 5.21 | 14.72 | 21.57 | |
| CAMP-tripletBackbone=Faster R-CNN, biGRU, Params=27.03M2023.08 | 5.12 | 12.89 | 21.12 | |
| AMFMN-softBackbone=ResNet18, biGRU, Params=35.94M2023.08 | 5.05 | 14.53 | 21.57 | |
| MTFNBackbone=Faster R-CNN, biGRU, Params=77.90M2023.08 | 5.02 | 12.52 | 19.74 | |
| LW-MCR-bBackbone=SqueezeNet, -, Params=1.65M2023.08 | 4.57 | 13.71 | 20.11 | |
| SCAN t-iBackbone=Faster R-CNN, biGRU, Params=13.68M2023.08 | 4.39 | 10.9 | 17.64 | |
| LW-MCR-uBackbone=SqueezeNet, -, Params=1.65M2023.08 | 4.39 | 13.35 | 20.29 | |
| CAMP-bceBackbone=Faster R-CNN, biGRU, Params=27.03M2023.08 | 4.2 | 10.24 | 15.45 | |
| VSE++Backbone=ResNet18, GRU, Params=15.78M2023.08 | 3.38 | 9.51 | 17.46 | |
| VSE++Paradigm=Supervised, Tuned on=RSICD2023.06 | 3.38 | 9.51 | 17.46 | |
| LW-MCR-dBackbone=SqueezeNet, -, Params=1.65M2023.08 | 3.29 | 12.52 | 19.93 | |
| LW-MCRParadigm=Supervised, Tuned on=RSICD2023.06 | 3.29 | 12.52 | 19.93 |