Pyramid Adversarial Training Improves ViT Performance

Open paper
Benchmark Model Rank Results
domain-generalization-on-imagenet-cPyramid Adversarial Training Improves ViT (Im21k)#9mean Corruption Error (mCE): 36.80Number of params: 87M
domain-generalization-on-imagenet-cPyramid Adversarial Training Improves ViT#15mean Corruption Error (mCE): 41.42
domain-generalization-on-imagenet-aPyramid Adversarial Training Improves ViT (Im21k)#12Top-1 accuracy %: 62.44
domain-generalization-on-imagenet-aPyramid Adversarial Training Improves ViT (384x384)#23Top-1 accuracy %: 36.41
domain-generalization-on-imagenet-sketchPyramid Adversarial Training Improves ViT (Im21k)#15Top-1 accuracy: 46.03
domain-generalization-on-imagenet-sketchPyramid Adversarial Training Improves ViT#19Top-1 accuracy: 41.04
domain-generalization-on-imagenet-rPyramid Adversarial Training Improves ViT (Im21k)#16Top-1 Error Rate: 42.16
domain-generalization-on-imagenet-rPyramid Adversarial Training Improves ViT#21Top-1 Error Rate: 46.08
image-classification-on-objectnetViT-B/16 (512x512) + Pyramid#35Top-1 Accuracy: 49.39
image-classification-on-objectnetViT-B/16 (512x512) + Pixel#37Top-1 Accuracy: 47.53
image-classification-on-objectnetViT-B/16 (512x512)#40Top-1 Accuracy: 46.68
image-classification-on-objectnetRegViT on 384x384 + Adv Pyramid#48Top-1 Accuracy: 39.79
image-classification-on-objectnetRegViT on 384x384 + Adv Pixel#52Top-1 Accuracy: 37.41
image-classification-on-objectnetRegViT on 384x384#56Top-1 Accuracy: 35.59
image-classification-on-objectnetRegViT on 384x384 + Random Pyramid#58Top-1 Accuracy: 34.83
image-classification-on-objectnetRegViT on 384x384 + Random Pixel#60Top-1 Accuracy: 34.12
image-classification-on-objectnetRegViT (RandAug) + Adv Pyramid#61Top-1 Accuracy: 32.92
image-classification-on-objectnetDiscrete ViT + Pixel#65Top-1 Accuracy: 30.98
image-classification-on-objectnetDiscrete ViT + Pyramid#66Top-1 Accuracy: 30.28
image-classification-on-objectnetRegViT (RandAug) + Adv Pixel#67Top-1 Accuracy: 30.11
image-classification-on-objectnetDiscrete ViT#68Top-1 Accuracy: 29.95
image-classification-on-objectnetRegViT (RandAug) + Random Pyramid#69Top-1 Accuracy: 29.41
image-classification-on-objectnetRegViT (RandAug)#70Top-1 Accuracy: 29.3
image-classification-on-objectnetRegViT (RandAug) + Random Pixel#73Top-1 Accuracy: 28.72
image-classification-on-objectnetMLP-Mixer + Pyramid#74Top-1 Accuracy: 28.6
image-classification-on-objectnetMLP-Mixer#79Top-1 Accuracy: 25.9
image-classification-on-objectnetViT + MixUp#81Top-1 Accuracy: 25.65
image-classification-on-objectnetMLP-Mixer + Pixel#83Top-1 Accuracy: 24.75
image-classification-on-objectnetViT + CutMix#86Top-1 Accuracy: 21.61
image-classification-on-objectnetViT#90Top-1 Accuracy: 17.36

Rank counts only results with a code link.