Image Classification on mini-ImageNet (test)
88.17Top-1 AccuracySpectral VMamba-B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Spectral VMamba-BFLOPs (G)=6.3G, Image Size=224x224, Model Scale=Base (B), Architecture Category=SSM-Based2025.03 | 88.17 | — | 97.27 | |
| Spectral VMamba-SFLOPs (G)=6.3G, Image Size=224x224, Model Scale=Small (S), Architecture Category=SSM-Based2025.03 | 88.09 | — | 97.08 | |
| Spectral VMamba-TFLOPs (G)=3.9G, Image Size=224x224, Model Scale=Tiny (T), Architecture Category=SSM-Based2025.03 | 87.86 | — | 97.25 | |
| VMamba-BFLOPs (G)=8.7G, Image Size=224x224, Model Scale=Base (B), Architecture Category=SSM-Based2025.03 | 87.17 | — | 96.96 | |
| XCiT-M24FLOPs (G)=16.2G, Image Size=224x224, Model Scale=Medium (M), Architecture Category=Transformer-Based2025.03 | 86.8 | — | 96.38 | |
| MSVMamba-TFLOPs (G)=4.6G, Image Size=224x224, Model Scale=Tiny (T), Architecture Category=SSM-Based2025.03 | 86.48 | — | 96.43 | |
| VMamba-SFLOPs (G)=8.7G, Image Size=224x224, Model Scale=Small (S), Architecture Category=SSM-Based2025.03 | 86.48 | — | 96.79 | |
| VMamba-TFLOPs (G)=4.9G, Image Size=224x224, Model Scale=Tiny (T), Architecture Category=SSM-Based2025.03 | 86.25 | — | 96.64 | |
| XCiT-S24FLOPs (G)=9.2G, Image Size=224x224, Model Scale=Small (S), Architecture Category=Transformer-Based2025.03 | 85.79 | — | 96.31 | |
| Swin-SFLOPs (G)=8.7G, Image Size=224x224, Model Scale=Small (S), Architecture Category=Transformer-Based2025.03 | 84.1 | — | 95.53 | |
| DeepInversionalpha=0.5, seed=12026.05 | 83.93 | — | — | |
| MSVMamba-MFLOPs (G)=1.5G, Image Size=224x224, Model Scale=Medium (M), Architecture Category=SSM-Based2025.03 | 83.72 | — | 95.81 | |
| Swin-TFLOPs (G)=4.5G, Image Size=224x224, Model Scale=Tiny (T), Architecture Category=Transformer-Based2025.03 | 83.25 | — | 95.54 | |
| DeepInversionalpha=0.5, seed=22026.05 | 82.93 | — | — | |
| Swin-BFLOPs (G)=15.4G, Image Size=224x224, Model Scale=Base (B), Architecture Category=Transformer-Based2025.03 | 82.77 | — | 95.33 | |
| FedFTGalpha=0.5, seed=12026.05 | 82.45 | — | — | |
| Co-Boostingalpha=0.5, seed=12026.05 | 82.43 | — | — | |
| DeepInversionalpha=0.5, seed=02026.05 | 82.42 | — | — | |
| FedMITRalpha=0.5, seed=12026.05 | 82.33 | — | — | |
| FedMITRalpha=0.5, seed=22026.05 | 82.27 | — | — | |
| DENSEalpha=0.5, seed=12026.05 | 82.24 | — | — | |
| FedFTGalpha=0.5, seed=22026.05 | 82.22 | — | — | |
| DENSEalpha=0.5, seed=22026.05 | 82.17 | — | — | |
| MSVMamba-NFLOPs (G)=0.9G, Image Size=224x224, Model Scale=Nano (N), Architecture Category=SSM-Based2025.03 | 82.16 | — | 95.1 | |
| LocalVim-TFLOPs (G)=1.5G, Image Size=224x224, Model Scale=Tiny (T), Architecture Category=SSM-Based2025.03 | 82.12 | — | 94.6 | |
| FedAvgalpha=0.5, seed=12026.05 | 82.12 | — | — | |
| Co-Boostingalpha=0.5, seed=22026.05 | 82.1 | — | — | |
| FedFTGalpha=0.5, seed=02026.05 | 82.09 | — | — | |
| Co-Boostingalpha=0.5, seed=02026.05 | 82.09 | — | — | |
| DENSEalpha=0.5, seed=02026.05 | 82.08 | — | — | |
| FedMITRalpha=0.5, seed=02026.05 | 81.93 | — | — | |
| FedAvgalpha=0.5, seed=02026.05 | 81.83 | — | — | |
| LocalVim-SFLOPs (G)=4.8G, Image Size=224x224, Model Scale=Small (S), Architecture Category=SSM-Based2025.03 | 81.68 | — | 93.63 | |
| FedAvgalpha=0.5, seed=22026.05 | 81.67 | — | — | |
| DeepInversionalpha=0.3, seed=12026.05 | 80.47 | — | — | |
| Vim-SFLOPs (G)=5.1G, Image Size=224x224, Model Scale=Small (S), Architecture Category=SSM-Based2025.03 | 79.7 | — | 93.2 | |
| Co-Boostingalpha=0.3, seed=12026.05 | 79.37 | — | — | |
| FedFTGalpha=0.3, seed=12026.05 | 79.32 | — | — | |
| DENSEalpha=0.3, seed=12026.05 | 79.28 | — | — | |
| FedMITRalpha=0.3, seed=12026.05 | 79.17 | — | — | |
| DeepInversionalpha=0.3, seed=02026.05 | 78.84 | — | — | |
| FedAvgalpha=0.3, seed=12026.05 | 78.81 | — | — | |
| DeepInversionalpha=0.3, seed=22026.05 | 77.88 | — | — | |
| FedFTGalpha=0.3, seed=22026.05 | 76.9 | — | — | |
| FedFTGalpha=0.3, seed=02026.05 | 76.69 | — | — | |
| FedMITRalpha=0.3, seed=02026.05 | 76.58 | — | — | |
| FedMITRalpha=0.3, seed=22026.05 | 76.56 | — | — | |
| Co-Boostingalpha=0.3, seed=22026.05 | 76.51 | — | — | |
| Co-Boostingalpha=0.3, seed=02026.05 | 76.49 | — | — | |
| DENSEalpha=0.3, seed=22026.05 | 76.48 | — | — | |
| Nat. Tra.Attack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 76.37 | — | — | |
| DENSEalpha=0.3, seed=02026.05 | 76.19 | — | — | |
| FedAvgalpha=0.3, seed=22026.05 | 75.92 | — | — | |
| FedAvgalpha=0.3, seed=02026.05 | 75.78 | — | — | |
| MAPEAttack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 74.86 | — | — | |
| Fully Supervised BaselineSupervision=Fully Supervised2026.03 | 73.95 | — | — | |
| Mix. Inf.Attack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 73.81 | — | — | |
| LDTAttack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 73.76 | — | — | |
| DeiT-BFLOPs (G)=17.5G, Image Size=224x224, Model Scale=Base (B), Architecture Category=Transformer-Based2025.03 | 72.43 | — | 90.14 | |
| HGDAttack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 72.41 | — | — | |
| DeiT-SFLOPs (G)=4.6G, Image Size=224x224, Model Scale=Small (S), Architecture Category=Transformer-Based2025.03 | 70.83 | — | 89.74 | |
| L2P-AHILBag Size=162026.03 | 70.26 | — | — | |
| FedMITRalpha=0.1, seed=02026.05 | 69.94 | — | — | |
| FedMITRalpha=0.1, seed=12026.05 | 68.67 | — | — | |
| MOIT+Noise Level=0%, Checkpoint=Best2020.12 | 68.28 | — | — | |
| MOIT+Noise Level=0%, Checkpoint=Last2020.12 | 67.82 | — | — | |
| Vim-TFLOPs (G)=1.5G, Image Size=224x224, Model Scale=Tiny (T), Architecture Category=SSM-Based2025.03 | 67.3 | — | 87.97 | |
| Pix. Def.Attack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 67.28 | — | — | |
| MOITNoise Level=0%, Checkpoint=Best2020.12 | 67.18 | — | — | |
| ROTBag Size=162026.03 | 67.02 | — | — | |
| LLP-DCBag Size=162026.03 | 66.9 | — | — | |
| SimPLENumber of labeled samples=4000, Backbone=WRN 28-2, K=72021.03 | 66.55 | — | — | |
| TVMAttack Type=Clean, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 66.22 | — | — | |
| FedMITRalpha=0.1, seed=22026.05 | 66.01 | — | — | |
| MAPEAttack Type=UPGD, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 65.88 | — | — | |
| MAPEAttack Type=BIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 65.43 | — | — | |
| MOIT+Noise Level=20%, Checkpoint=Best2020.12 | 64.98 | — | — | |
| MOITNoise Level=20%, Checkpoint=Best2020.12 | 64.82 | — | — | |
| MOITNoise Level=0%, Checkpoint=Last2020.12 | 64.72 | — | — | |
| MAPEAttack Type=FGSM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 64.58 | — | — | |
| DLLPBag Size=162026.03 | 64.53 | — | — | |
| MAPEAttack Type=VNIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 64.26 | — | — | |
| LLP-VATBag Size=162026.03 | 64.17 | — | — | |
| Pix. Def.Attack Type=BIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 63.5 | — | — | |
| MOITNoise Level=20%, Checkpoint=Last2020.12 | 63.14 | — | — | |
| ELRNoise Level=0%, Checkpoint=Best2020.12 | 63.12 | — | — | |
| MOIT+Noise Level=20%, Checkpoint=Last2020.12 | 63.1 | — | — | |
| TVMAttack Type=BIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 62.9 | — | — | |
| DeepInversionalpha=0.1, seed=02026.05 | 62.71 | — | — | |
| LDTAttack Type=UPGD, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 62.65 | — | — | |
| Mix. Inf.Attack Type=BIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 62.52 | — | — | |
| MOIT+Noise Level=40%, Checkpoint=Best2020.12 | 62.36 | — | — | |
| FedMITRalpha=0.05, seed=12026.05 | 62.36 | — | — | |
| HGDAttack Type=UPGD, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 62.31 | — | — | |
| FedMITRalpha=0.05, seed=02026.05 | 62.26 | — | — | |
| HGDAttack Type=BIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 62.25 | — | — | |
| Pix. Def.Attack Type=UPGD, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 62.08 | — | — | |
| LDTAttack Type=BIM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 61.97 | — | — | |
| Mix. Inf.Attack Type=FGSM, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 61.89 | — | — | |
| Mix. Inf.Attack Type=UPGD, Substitute Model=ViT-S/16, Target Model=ResNet-342026.06 | 61.89 | — | — |