Image Classification on ImageNet (10%)
90.06Top-1 AccVisformer-S
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Visformer-SFLOPS (G)=4.9, Parameters (M)=40.22021.04 | 90.06 | — | |
| ResNet-50FLOPS (G)=4.1, Parameters (M)=25.62021.04 | 89.9 | — | |
| Meta Co-TrainingBackbone=MLP, Trained on large model embeddings=true, Protocol=Meta Co-Training2023.11 | 85.8 | — | |
| REACTBackbone=ViT-L, Protocol=REACT2023.11 | 85.1 | — | |
| Co-TrainingBackbone=MLP, Trained on large model embeddings=true, Protocol=Co-Training2023.11 | 85.1 | — | |
| CLIPBackbone=ViT-L, Protocol=Fine-tuned2023.11 | 84.7 | — | |
| Deep EnsembleTrained on large model embeddings=true, Protocol=Deep Ensemble2023.11 | 84.3 | — | |
| Semi-ViTBackbone=ViT-L, Protocol=Self-trained2023.11 | 83.3 | — | |
| DINOv2Backbone=ViT-L, Protocol=Linear2023.11 | 82.9 | — | |
| SimCLR v2Arch.=RN152w3+SK, Param.=795M2021.03 | 80.1 | — | |
| DeiT-SFLOPS (G)=4.6, Parameters (M)=21.82021.04 | 80.06 | — | |
| SimCLRv2Backbone=ResNet152-w2, Protocol=Fine-tuned2023.11 | 79.4 | — | |
| SEERArch.=RG256, Param.=1.3B2021.03 | 77.9 | — | |
| BYOLArch.=RN200, Param.=250M2021.03 | 77.7 | — | |
| SEERArch.=RG128, Param.=693M2021.03 | 76.7 | — | |
| EsViTBackbone=Swin-B, W=14, Protocol=Linear2023.11 | 74.4 | — | |
| CowMixArch.=RN152, Param.=265M2021.03 | 73.9 | — | |
| MPLBackbone=EfficientNet-B6-Wide, Protocol=MPL2023.11 | 73.9 | — | |
| FixMatchArch.=RN50, Param.=24M2021.03 | 71.5 | — | |
| FixMatchBackbone=ResNet-50, Protocol=FixMatch2023.11 | 71.46 | — | |
| SwAVArch.=RN50, Param.=24M2021.03 | 70.2 | — | |
| SwAVBackbone=ResNet-50, Evaluation Protocol=Fine-tuning, multi-crop=true2021.04 | 70.2 | 89.9 | |
| SWAVBackbone=RN50-w4, Protocol=Fine-tuned2023.11 | 70.2 | — | |
| NNCLRBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | 69.8 | 89.3 | |
| BYOLBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | 68.8 | 89 | |
| UDABackbone=ResNet50, Protocol=UDA2023.11 | 68.78 | — | |
| SimCLRArch.=RN50, Param.=24M2021.03 | 65.6 | — | |
| SimCLRBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | 65.6 | 87.8 | |
| SupervisedBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | 56.4 | 80.4 | |
| ScratchArch.=RG128, Param.=848M2021.03 | 54.5 | — | |
| Deep Co-TrainingBackbone=ResNet-18, Protocol=Co-Training2023.11 | 53.5 | — | |
| MAEBackbone=ViT-L, Protocol=MLP2023.11 | 48.5 | — | |
| InstDiscBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | — | 77.4 | |
| PCLBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | — | 86.2 | |
| PIRLBackbone=ResNet-50, Evaluation Protocol=Fine-tuning2021.04 | — | 83.8 |