Image Classification on ImageNet 1% labeled subset 1k (test)
93.2Top-1 AccResNet-50
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ResNet-50Training Setting=Elite, Epochs=3002021.04 | 93.2 | — | |
| Visformer-STraining Setting=Elite, Epochs=3002021.04 | 91.6 | — | |
| ResNet-50-55%Training Setting=Elite, Epochs=3002021.04 | 91.4 | — | |
| Visformer-TiTraining Setting=Elite, Epochs=3002021.04 | 90.6 | — | |
| DeiT-TiTraining Setting=Elite, Epochs=3002021.04 | 74.4 | — | |
| DeiT-STraining Setting=Elite, Epochs=3002021.04 | 73.4 | — | |
| WCLLearning Paradigm=Self-Supervised, Fine-tuning Layer=Projection Head2021.10 | 65 | 86.3 | |
| FNCancelLearning Paradigm=Self-Supervised, Fine-tuning Layer=Projection Head2021.10 | 63.7 | 85.3 | |
| SimCLR v2Learning Paradigm=Self-Supervised, Fine-tuning Layer=Projection Head, Backbone=ResNet-1012021.10 | 62.1 | — | |
| WCLLearning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | 58.3 | 79.9 | |
| SimCLR v2Learning Paradigm=Self-Supervised, Fine-tuning Layer=Projection Head, Backbone=ResNet-502021.10 | 57.9 | — | |
| SwAVLearning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | 53.9 | 78.5 | |
| BYOLLearning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | 53.2 | 78.4 | |
| SimCLR v1Learning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | 48.3 | 75.5 | |
| PIRLLearning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | 30.7 | 60.4 | |
| SupervisedLearning Paradigm=Supervised2021.10 | 25.4 | 56.4 | |
| Visformer-STraining Setting=Elite, Epochs=3002021.04 | 16.56 | — | |
| ResNet-50Training Setting=Elite, Epochs=3002021.04 | 13.59 | — | |
| Visformer-TiTraining Setting=Elite, Epochs=3002021.04 | 11.79 | — | |
| ResNet-50-55%Training Setting=Elite, Epochs=3002021.04 | 10.68 | — | |
| DeiT-STraining Setting=Elite, Epochs=3002021.04 | 6.94 | — | |
| DeiT-TiTraining Setting=Elite, Epochs=3002021.04 | 6.53 | — | |
| InstDiscLearning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | — | 39.2 | |
| PCLLearning Paradigm=Self-Supervised, Fine-tuning Layer=AvgPool2021.10 | — | 75.6 | |
| S4LLearning Paradigm=Semi-supervised2021.10 | — | 53.4 | |
| UDALearning Paradigm=Semi-supervised2021.10 | — | 68.8 |