Understanding The Robustness in Vision Transformers

Open paper
Benchmark Model Rank Results
domain-generalization-on-imagenet-aFAN-Hybrid-L(IN-21K, 384)#6Top-1 accuracy %: 74.5
domain-generalization-on-imagenet-cFAN-L-Hybrid (IN-22k)#8mean Corruption Error (mCE): 35.8Top 1 Accuracy: 73.6…
domain-generalization-on-imagenet-cFAN-B-Hybrid (IN-22k)#14mean Corruption Error (mCE): 41.0Top 1 Accuracy: 70.5…
domain-generalization-on-imagenet-cFAN-L-Hybrid#20mean Corruption Error (mCE): 43.0Top 1 Accuracy: 67.7…
domain-generalization-on-imagenet-rFAN-Hybrid-L(IN-21K, 384))#3Top-1 Error Rate: 28.9
image-classification-on-imagenetFAN-L-Hybrid++#105Top 1 Accuracy: 87.1%Number of params: 76.8M
object-detection-on-coco-minivalFAN-L-Hybrid#51box AP: 55.1
semantic-segmentation-on-cityscapes-valFAN-L-Hybrid#33mIoU: 82.3
semantic-segmentation-on-densepassFAN (MiT-B1)#11mIoU: 42.54%

Rank counts only results with a code link.