Image Classification on STL-10 100 labels
7.1Top-1 Error RateSoftMatch/FreeMatch + SemiReward
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SoftMatch/FreeMatch + SemiRewardBackbone=ViT [Dosovitskiy et al., 2021] / ResNet-50 [He et al., 2016], SemiReward=true2023.10 | 7.1 | 0.3 | |
| FlexMatchBackbone=ViT [Dosovitskiy et al., 2021] / ResNet-50 [He et al., 2016], SemiReward=false2023.10 | 7.13 | — | |
| FlexMatch + SemiRewardBackbone=ViT [Dosovitskiy et al., 2021] / ResNet-50 [He et al., 2016], SemiReward=true2023.10 | 7.59 | 0.3 | |
| SoftMatch/FreeMatchBackbone=ViT [Dosovitskiy et al., 2021] / ResNet-50 [He et al., 2016], SemiReward=false2023.10 | 8.1 | — | |
| Pseudo Label + SemiRewardBackbone=ViT [Dosovitskiy et al., 2021] / ResNet-50 [He et al., 2016], SemiReward=true2023.10 | 10.88 | 0.3 | |
| Pseudo LabelBackbone=ViT [Dosovitskiy et al., 2021] / ResNet-50 [He et al., 2016], SemiReward=false2023.10 | 11.25 | — |