Big Self-Supervised Models are Strong Semi-Supervised Learners

Open paper
Benchmark Model Rank Results
self-supervised-image-classification-on-imagenetSimCLRv2 (ResNet-152 x3, SK)#25Top 1 Accuracy: 79.8%Top 5 Accuracy: 94.9%Number of Params: 795M
self-supervised-image-classification-on-imagenetSimCLRv2 (ResNet-50 x2)#63Top 1 Accuracy: 75.6%Top 5 Accuracy: 92.7%Number of Params: 94M
self-supervised-image-classification-on-imagenetSimCLRv2 (ResNet-50)#90Top 1 Accuracy: 71.7%Top 5 Accuracy: 90.4%Number of Params: 24M
self-supervised-image-classification-on-imagenet-finetunedSimCLRv2 (ResNet-152, 3×+SK)#46Top 1 Accuracy: 83.1%Number of Params: 795M
semi-supervised-image-classification-on-imagenet-1-labeled-dataSimCLRv2 self-distilled (ResNet-152 x3, SK)#7Top 1 Accuracy: 76.6%Top 5 Accuracy: 93.4%
semi-supervised-image-classification-on-imagenet-1-labeled-dataSimCLRv2 distilled (ResNet-50 x2, SK)#8Top 1 Accuracy: 75.9%Top 5 Accuracy: 93.0%
semi-supervised-image-classification-on-imagenet-1-labeled-dataSimCLRv2 (ResNet-152 x3, SK)#10Top 1 Accuracy: 74.9%Top 5 Accuracy: 92.3%
semi-supervised-image-classification-on-imagenet-1-labeled-dataSimCLRv2 distilled (ResNet-50)#11Top 1 Accuracy: 73.9%Top 5 Accuracy: 91.5%
semi-supervised-image-classification-on-imagenet-1-labeled-dataSimCLRv2 (ResNet-50 ×2)#25Top 1 Accuracy: 66.3%Top 5 Accuracy: 87.4%
semi-supervised-image-classification-on-imagenet-1-labeled-dataSimCLRv2 (ResNet-50)#37Top 1 Accuracy: 57.9%Top 5 Accuracy: 82.5%
semi-supervised-image-classification-on-imagenet-10-labeled-dataSimCLRv2 self-distilled (ResNet-152 x3, SK)#7Top 1 Accuracy: 80.9%Top 5 Accuracy: 95.5%
semi-supervised-image-classification-on-imagenet-10-labeled-dataSimCLRv2 distilled (ResNet-50 x2, SK)#8Top 1 Accuracy: 80.2%Top 5 Accuracy: 95.0%
semi-supervised-image-classification-on-imagenet-10-labeled-dataSimCLRv2 (ResNet-152 x3, SK)#9Top 1 Accuracy: 80.1%Top 5 Accuracy: 95.0%
semi-supervised-image-classification-on-imagenet-10-labeled-dataSimCLRv2 distilled (ResNet-50)#15Top 1 Accuracy: 77.5%Top 5 Accuracy: 93.4%
semi-supervised-image-classification-on-imagenet-10-labeled-dataSimCLRv2 (ResNet-50 x2)#28Top 1 Accuracy: 73.9%Top 5 Accuracy: 91.9%
semi-supervised-image-classification-on-imagenet-10-labeled-dataSimCLRv2 (ResNet-50)#38Top 1 Accuracy: 68.4%Top 5 Accuracy: 89.2%

Rank counts only results with a code link.