Self-training with Noisy Student improves ImageNet classification

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetNoisyStudent (EfficientNet-L2)#46Top 1 Accuracy: 88.4%Number of params: 480MHardware Burden: 51800G
image-classification-on-imagenetNoisyStudent (EfficientNet-B7)#112Top 1 Accuracy: 86.9%Number of params: 66MGFLOPs: 37
image-classification-on-imagenetNoisyStudent (EfficientNet-B6)#141Top 1 Accuracy: 86.4%Number of params: 43M
image-classification-on-imagenetNoisyStudent (EfficientNet-B5)#167Top 1 Accuracy: 86.1%Number of params: 30M
image-classification-on-imagenetNoisyStudent (EfficientNet-B4)#233Top 1 Accuracy: 85.3%Number of params: 19M
image-classification-on-imagenetNoisyStudent (EfficientNet-B3)#336Top 1 Accuracy: 84.1%Number of params: 12M
image-classification-on-imagenetNoisyStudent (EfficientNet-B2)#519Top 1 Accuracy: 82.4%Number of params: 9.2M
image-classification-on-imagenetNoisyStudent (EfficientNet-B1)#606Top 1 Accuracy: 81.5%Number of params: 7.8M
image-classification-on-imagenetNoisyStudent (EfficientNet-B0)#759Top 1 Accuracy: 78.8%Number of params: 5.3M
image-classification-on-imagenet-realEfficientNet-L2#16Accuracy: 90.55%Params: 480M

Rank counts only results with a code link.