Image Classification on ImageNet-S 1.0 (test)
88Top-1 AccuracySwin-B
Evaluation results
| Method | Links | |
|---|---|---|
| Swin-BFine-tuned with ImageNet-S semi-supervised set=false2021.06 | 88 | |
| Swin-SFine-tuned with ImageNet-S semi-supervised set=false2021.06 | 87.8 | |
| Res2Net-101Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 85.6 | |
| ResNeXt-101Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 85.5 | |
| EfficientNet-B3Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 85.3 | |
| Res2Net-50Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 84.8 | |
| ResNeXt-50Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 84.4 | |
| ResNet-101Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 84.3 | |
| DenseNet-161Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 84.3 | |
| ResNet-50Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 83.6 | |
| Inception V3Fine-tuned with ImageNet-S semi-supervised set=false2021.06 | 77.7 | |
| Federated Averagingdataset skew (s)=0, synchronization steps (u)=200, training steps=60k2022.11 | 70.5 | |
| Federated Averagingdataset skew (s)=0, synchronization steps (u)=1000, training steps=60k2022.11 | 69.1 | |
| Supervisedtraining steps=60k, data scope=90% of ImageNet2022.11 | 68.9 | |
| MHD+dataset skew (s)=0, training steps=180k, public dataset scope=entire ImageNet2022.11 | 68.6 | |
| MHDdataset skew (s)=0, training steps=60k, public dataset scope=90% of ImageNet2022.11 | 59.9 | |
| Separatedataset skew (s)=02022.11 | 46.3 |