Image Classification on CUB 200 2011
97.1AccuracyViT-L/16
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ViT-L/16N-shot=10, Probe=linear regression2023.05 | 97.1 | — | — | |
| VT-FSLLearning Paradigm=Training-based2026.03 | 90.98 | — | — | |
| SARELearning Paradigm=Training-based2026.03 | 90.76 | — | — | |
| OSCS-SupConBackbone=TinyViT (21M)2026.05 | 90.5 | — | — | |
| SCS-SupConBackbone=TinyViT (21M)2025.12 | 90.2 | — | — | |
| CS-SupConBackbone=TinyViT (21M)2025.12 | 89.8 | — | — | |
| CS-SupCon w. ov.Backbone=TinyViT (21M)2025.12 | 89.8 | — | — | |
| CS-SupConBackbone=TinyViT (21M)2026.05 | 89.8 | — | — | |
| CS-SupCon w. ov.Backbone=TinyViT (21M)2026.05 | 89.8 | — | — | |
| SSF*Backbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.39, Data Augmentation=Advanced (cutmix, mixup, label smoothing)2023.10 | 89.5 | — | — | |
| DACLBackbone=TinyViT (21M)2025.12 | 89.5 | — | — | |
| SGD-DualOpt_Fine-tuneBackbone=ConvNeXt-B2026.04 | 89.47 | — | — | |
| CSTCNBackbone=TinyViT (21M)2025.12 | 89.4 | — | — | |
| TimeSCLBackbone=TinyViT (21M)2025.12 | 89.4 | — | — | |
| ARC*Backbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.25, Data Augmentation=Advanced (cutmix, mixup, label smoothing)2023.10 | 89.3 | — | — | |
| PCLBackbone=TinyViT (21M)2025.12 | 89.3 | — | — | |
| FNCLBackbone=TinyViT (21M)2025.12 | 89.3 | — | — | |
| PaCoBackbone=TinyViT (21M)2025.12 | 89.2 | — | — | |
| SupConBackbone=TinyViT (21M)2025.12 | 89.1 | — | — | |
| CSA-RSICBackbone=TinyViT (21M)2025.12 | 89.1 | — | — | |
| SupConBackbone=TinyViT (21M)2026.05 | 89.1 | — | — | |
| L2-SPBackbone=ConvNeXt-B2026.04 | 88.82 | — | — | |
| SoViT-400m/14N-shot=10, Probe=linear regression2023.05 | 88.8 | — | — | |
| FixResBackbone=SENet-1542019.06 | 88.7 | — | — | |
| MPN-COV2019.06 | 88.7 | — | — | |
| BaselineBackbone=TinyViT (21M)2025.12 | 88.7 | — | — | |
| Circle LossBackbone=TinyViT (21M)2025.12 | 88.7 | — | — | |
| BaselineBackbone=TinyViT (21M)2026.05 | 88.7 | — | — | |
| FullBackbone=ConvNeXt-B2026.04 | 88.59 | — | — | |
| VPT-DeepBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.852023.10 | 88.5 | — | — | |
| ARCBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.252023.10 | 88.5 | — | — | |
| ViT-g/14N-shot=10, Probe=linear regression2023.05 | 88.5 | — | — | |
| SENet-154Evaluation Protocol=Baseline2019.06 | 88.4 | — | — | |
| BiasBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.282023.10 | 88.4 | — | — | |
| ARCattBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.222023.10 | 88.4 | — | — | |
| LoRABackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.442023.10 | 88.3 | — | — | |
| VPTBackbone=ConvNeXt-B2026.04 | 87.88 | — | — | |
| LinearBackbone=ConvNeXt-B2026.04 | 87.85 | — | — | |
| GeoProtoBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 87.8 | — | — | |
| Full fine-tuningBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=85.982023.10 | 87.3 | — | — | |
| AdapterBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.412023.10 | 87.1 | — | — | |
| SVD-ViTComponents=SPC + SSVA + ID-RSVD, Layer=102026.02 | 87.02 | — | — | |
| SVD-ViTComponents=SPC, Layer=112026.02 | 86.74 | — | — | |
| VPT-ShallowBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.252023.10 | 86.7 | — | — | |
| SVD-ViTComponents=SPC + SSVA, Layer=102026.02 | 86.54 | — | — | |
| SVD-ViTComponents=SPC + SSVA, Layer=92026.02 | 86.47 | — | — | |
| SVD-ViTComponents=SPC + ID-RSVD, Layer=112026.02 | 86.45 | — | — | |
| Deformable ProtoPNetPrototype size=2 x 2, Deformation=False (nd), Backbone=ResNet-50, Pre-training=iNaturalist [41]2021.11 | 86.4 | — | — | |
| Deformable ProtoPNetPrototype size=2 x 2, Deformation=True, Backbone=ResNet-50, Pre-training=iNaturalist [41]2021.11 | 86.4 | — | — | |
| SVD-ViTComponents=SPC + SSVA, Layer=112026.02 | 86.37 | — | — | |
| SVD-ViTComponents=SPC + ID-RSVD, Layer=102026.02 | 86.33 | — | — | |
| SVD-ViTComponents=SPC + SSVA + ID-RSVD, Layer=112026.02 | 86.3 | — | — | |
| MGProtoBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 86.2 | — | — | |
| SVD-ViTComponents=SPC + SSVA + ID-RSVD, Layer=92026.02 | 86.18 | — | — | |
| LoRABackbone=ConvNeXt-B2026.04 | 86.16 | — | — | |
| Deformable ProtoPNetPrototype size=3 x 3, Deformation=True, Backbone=ResNet-50, Pre-training=iNaturalist [41]2021.11 | 86.1 | — | — | |
| Deformable ProtoPNetPrototype size=3 x 3, Deformation=False (nd), Backbone=ResNet-50, Pre-training=iNaturalist [41]2021.11 | 85.9 | — | — | |
| SVD-ViTComponents=SPC, Layer=92026.02 | 85.85 | — | — | |
| ResNet-50Features=iNaturalist, Training set=CUB2022.07 | 85.83 | — | — | |
| SVD-ViTComponents=SPC, Layer=102026.02 | 85.61 | — | — | |
| SDFA-SABackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 85.6 | — | — | |
| CBCBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 85.5 | — | — | |
| kNNFeatures=iNaturalist, Training set=CUB2022.07 | 85.46 | — | — | |
| SVD-ViTComponents=SPC + ID-RSVD, Layer=92026.02 | 85.4 | — | — | |
| TesNetBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 85.4 | — | — | |
| Linear probingBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.182023.10 | 85.3 | — | — | |
| PS-CBMBackbone=CLIP ViT-L/142025.11 | 85.3 | 64.1 | — | |
| ProtoArgNetBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 85.1 | — | — | |
| EMD-CorrFeatures=iNaturalist, Training set=CUB2022.07 | 84.98 | — | — | |
| Adam-DualOpt_Fine-tuneBackbone=ViT-B2026.04 | 84.84 | — | — | |
| IndividualBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Number of Merged Models=12024.05 | 84.79 | — | — | |
| ViTCLS=8, Layer=-2026.02 | 84.64 | — | — | |
| Linear ProbeBackbone=CLIP ViT-L/142025.11 | 84.5 | — | — | |
| VLG-CBMBackbone=CLIP ViT-L/142025.11 | 84.5 | 63.4 | — | |
| ViTCLS=1, Layer=-2026.02 | 84.5 | — | — | |
| GeoProtoBackbone=DenseNet-161, Pre-trained=ImageNet2025.09 | 84.3 | — | — | |
| ST-ProtoPNetBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 83.9 | — | — | |
| DN-CBMBackbone=CLIP ViT-L/142025.11 | 83.3 | 58.2 | — | |
| CHM-CorrFeatures=iNaturalist, Training set=CUB2022.07 | 83.27 | — | — | |
| SSFBackbone=ViT-B/16, Pre-trained=ImageNet-21k, Params(M)=0.39, Data Augmentation=Basic2023.10 | 82.7 | — | — | |
| ProtoPoolBackbone=ResNet-50, Pre-trained=iNaturalist2025.09 | 82.6 | — | — | |
| CBCBackbone=DenseNet-161, Pre-trained=ImageNet2025.09 | 82.3 | — | — | |
| MGProtoBackbone=DenseNet-161, Pre-trained=ImageNet2025.09 | 82.1 | — | — | |
| GeoProtoBackbone=ResNet-34, Pre-trained=ImageNet2025.09 | 82.1 | — | — | |
| LaBoBackbone=CLIP ViT-L/142025.11 | 82 | 56.5 | — | |
| VPTBackbone=ViT-B2026.04 | 81.86 | — | — | |
| L2-SPBackbone=ViT-B2026.04 | 81.65 | — | — | |
| CHM-Corr+Features=iNaturalist, Training set=CUB2022.07 | 81.54 | — | — | |
| ProtoArgNetBackbone=DenseNet-161, Pre-trained=ImageNet2025.09 | 81.5 | — | — | |
| FullBackbone=ViT-B2026.04 | 81.34 | — | — | |
| Deformable ProtoPNetPrototype size=2 x 2, Deformation=True, Backbone=DenseNet-1612021.11 | 81.2 | — | — | |
| GeoProtoBackbone=DenseNet-121, Pre-trained=ImageNet2025.09 | 81.2 | — | — | |
| ProtoPNetBackbone=ResNet-50, Pre-training=iNaturalist [41], Retrained on full images=True2021.11 | 81.1 | — | — | |
| Deformable ProtoPNetPrototype size=2 x 2, Deformation=False (nd), Backbone=DenseNet-1612021.11 | 80.8 | — | — | |
| V2C-CBMBackbone=CLIP ViT-L/142025.11 | 80.8 | 55.7 | — | |
| DCBMBackbone=CLIP ViT-L/142025.11 | 80.8 | 58.4 | — | |
| MCPNetBackbone=ResNet152, Explanation=Multi-Scale2024.04 | 80.79 | — | — | |
| BYOLBackbone=TinyViT (21M)2025.12 | 80.7 | — | — | |
| BYOLBackbone=TinyViT (21M)2026.05 | 80.7 | — | — | |
| ProtoKNNBackbone=DenseNet-161, Pre-trained=ImageNet2025.09 | 80.6 | — | — |