Image Classification on CIFAR-100 (Top-1 Accuracy, Top-5 Accuracy)
82.9Top-1 AccDINO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DINOEvaluation Protocol=Linear Eval., Backbone=ViT-Base/16, Pre-trained=ImageNet-1K2022.04 | 82.9 | — | |
| MSNEvaluation Protocol=Linear Eval., Backbone=ViT-Base/16, Pre-trained=ImageNet-1K, Masking ratio=0.32022.04 | 82.8 | — | |
| NNCLRProtocol=linear evaluation2022.08 | 79 | — | |
| RényiCLProtocol=linear evaluation2022.08 | 79 | — | |
| BYOLProtocol=linear evaluation2022.08 | 78.4 | — | |
| SupervisedProtocol=linear evaluation2022.08 | 78.3 | — | |
| SimCLRProtocol=linear evaluation2022.08 | 71.6 | — | |
| EACNoise Rate=10%, Backbone=ResNet-182022.07 | 70.93 | 90.15 | |
| EACNoise Rate=20%, Backbone=ResNet-182022.07 | 66.73 | 87.01 | |
| SCNNoise Rate=10%, Backbone=ResNet-182022.07 | 65.18 | 86.6 | |
| BaselineNoise Rate=10%, Backbone=ResNet-182022.07 | 64.56 | 85.37 | |
| EACNoise Rate=30%, Backbone=ResNet-182022.07 | 60.59 | 82.84 | |
| SCNNoise Rate=20%, Backbone=ResNet-182022.07 | 60.38 | 82.11 | |
| BaselineNoise Rate=20%, Backbone=ResNet-182022.07 | 57.33 | 78.93 | |
| SCNNoise Rate=30%, Backbone=ResNet-182022.07 | 56.19 | 78.3 | |
| BaselineNoise Rate=30%, Backbone=ResNet-182022.07 | 49.7 | 72.55 | |
| RefProtoFLalpha=1002026.01 | 20.02 | — | |
| FedProtoalpha=1002026.01 | 18.42 | — | |
| FedNovaalpha=1002026.01 | 17.22 | — | |
| FedAvgalpha=1002026.01 | 16.75 | — | |
| RefProtoFLalpha=0.52026.01 | 16.5 | — | |
| FedProtoalpha=0.52026.01 | 16.28 | — | |
| VHLalpha=1002026.01 | 15.57 | — | |
| FedAvgalpha=0.52026.01 | 14.56 | — | |
| FedGHalpha=1002026.01 | 14.55 | — | |
| FedNovaalpha=0.52026.01 | 14.46 | — | |
| FedGHalpha=0.52026.01 | 14.16 | — | |
| VHLalpha=0.52026.01 | 13.24 | — | |
| FedProxalpha=0.52026.01 | 12.81 | — | |
| FASTalpha=1002026.01 | 12.13 | — | |
| FASTalpha=0.52026.01 | 9.97 | — | |
| FedProxalpha=1002026.01 | 8.74 | — |