Linear Classification on ImageNet official (val)
76AccuracySupervised ResNet50 v2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Supervised ResNet50 v2Family=Residual, Implementation=Previous2019.01 | 76 | — | |
| Supervised ResNet50 v2Family=Residual, Implementation=This Paper2019.01 | 75.8 | — | |
| Supervised VGG19Family=VGG-style, Implementation=This Paper2019.01 | 75 | — | |
| Supervised RevNet50Family=Residual, Implementation=Previous2019.01 | 74.8 | — | |
| Supervised RevNet50Family=Residual, Implementation=This Paper2019.01 | 74.4 | — | |
| Supervised VGG19Family=VGG-style, Implementation=Previous2019.01 | 72.7 | — | |
| RotationFamily=AlexNet, Implementation=This Paper2019.01 | 55.4 | — | |
| Rel. Patch Loc.Family=Residual, Implementation=This Paper2019.01 | 51.4 | — | |
| CPCFamily=Residual, Implementation=Previous2019.01 | 48.7 | — | |
| ExemplarFamily=Residual, Implementation=This Paper2019.01 | 46 | — | |
| JigsawFamily=AlexNet, Implementation=This Paper2019.01 | 44.6 | — | |
| DeepClusteringFamily=VGG-style, Implementation=Previous2019.01 | 41 | — | |
| RotationFamily=AlexNet, Implementation=Previous2019.01 | 38.7 | — | |
| CC+vgg-Jigsaw++Family=VGG-style, Implementation=Previous2019.01 | 37.3 | — | |
| Rel. Patch Loc.Family=Residual, Implementation=Previous2019.01 | 36.2 | — | |
| Split-BrainFamily=AlexNet, Implementation=Previous2019.01 | 35.4 | — | |
| JigsawFamily=AlexNet, Implementation=Previous2019.01 | 34.7 | — | |
| CountingFamily=AlexNet, Implementation=Previous2019.01 | 34.3 | — | |
| ExemplarFamily=Residual, Implementation=Previous2019.01 | 31.5 | — | |
| BYOL (repro.)Base Algorithm=BYOL, Augmentation=Standard, Reproduction=true2021.10 | — | 67.8 | |
| BYOL w/ DSSLBase Algorithm=BYOL, Method=DSSL, Augmentation=Standard + Heavy (T~)2021.10 | — | 68.3 | |
| BYOL w/ T~Base Algorithm=BYOL, Augmentation=Standard + Heavy (T~)2021.10 | — | 68.3 | |
| SimCLR (repro.)Base Algorithm=SimCLR, Augmentation=Standard, Reproduction=true2021.10 | — | 59.5 | |
| SimCLR w/ DSSLBase Algorithm=SimCLR, Method=DSSL, Augmentation=Standard + Heavy (T~)2021.10 | — | 60.3 | |
| SimCLR w/ T~Base Algorithm=SimCLR, Augmentation=Standard + Heavy (T~)2021.10 | — | 60 | |
| SimSiam (repro.)Base Algorithm=SimSiam, Augmentation=Standard, Reproduction=true2021.10 | — | 67.4 | |
| SimSiam w/ DSSLBase Algorithm=SimSiam, Method=DSSL, Augmentation=Standard + Heavy (T~)2021.10 | — | 68.6 |