Fixing the train-test resolution discrepancy: FixEfficientNet

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetFixEfficientNet-L2#41Top 1 Accuracy: 88.5%Number of params: 480MGFLOPs: 585
image-classification-on-imagenetFixEfficientNet-B7#100Top 1 Accuracy: 87.1%Number of params: 66MGFLOPs: 82
image-classification-on-imagenetFixEfficientNet-B6#123Top 1 Accuracy: 86.7%Number of params: 43M
image-classification-on-imagenetFixEfficientNet-B5#142Top 1 Accuracy: 86.4%Number of params: 30M
image-classification-on-imagenetFixEfficientNet-B4#180Top 1 Accuracy: 85.9%Number of params: 19M
image-classification-on-imagenetFixEfficientNet-B8#199Top 1 Accuracy: 85.7%
image-classification-on-imagenetFixEfficientNet-B3#257Top 1 Accuracy: 85%Number of params: 12M
image-classification-on-imagenetFixEfficientNetB4#351Top 1 Accuracy: 84.0%Number of params: 19M
image-classification-on-imagenetFixEfficientNet-B2#396Top 1 Accuracy: 83.6%Number of params: 9.2M
image-classification-on-imagenetFixEfficientNet-B1#504Top 1 Accuracy: 82.6%Number of params: 7.8M
image-classification-on-imagenetFixEfficientNet-B0#678Top 1 Accuracy: 80.2%Number of params: 5.3MGFLOPs: 1.60
image-classification-on-imagenet-realFixEfficientNet-L2#9Accuracy: 90.9%Params: 480M
image-classification-on-imagenet-realFixEfficientNet-B8#21Accuracy: 90.0%Params: 87M

Rank counts only results with a code link.