List your company
ResearchFormerly WizWand

Image Classification on CIFAR-100 resized to 224x224 (test)

88.4AccuracyBEiT large
66.5672.2377.983.57May 21, 2022
Updated 8mo ago

Evaluation results

MethodLinks
BEiT largeepsilon (ε)=8, pretrained_dataset=ImageNet21k, parameters=303M, clipping_method=mixed ghost clipping2022.0588.4
BEiT largeepsilon (ε)=4, pretrained_dataset=ImageNet21k, parameters=303M, clipping_method=mixed ghost clipping2022.0587.7
BEiT largeepsilon (ε)=2, pretrained_dataset=ImageNet21k, parameters=303M, clipping_method=mixed ghost clipping2022.0586.2
BEiT largeepsilon (ε)=1, pretrained_dataset=ImageNet21k, parameters=303M, clipping_method=mixed ghost clipping2022.0583
De et al.epsilon (ε)=8, pretrained_dataset=ImageNet1k2022.0581.8
De et al.epsilon (ε)=4, pretrained_dataset=ImageNet1k2022.0579.2
CrossViT baseepsilon (ε)=8, pretrained_dataset=ImageNet1k, parameters=104M, clipping_method=mixed ghost clipping2022.0577.8
CrossViT baseepsilon (ε)=4, pretrained_dataset=ImageNet1k, parameters=104M, clipping_method=mixed ghost clipping2022.0576.7
De et al.epsilon (ε)=2, pretrained_dataset=ImageNet1k2022.0574.7
CrossViT baseepsilon (ε)=2, pretrained_dataset=ImageNet1k, parameters=104M, clipping_method=mixed ghost clipping2022.0574.3
CrossViT baseepsilon (ε)=1, pretrained_dataset=ImageNet1k, parameters=104M, clipping_method=mixed ghost clipping2022.0571.9
De et al.epsilon (ε)=1, pretrained_dataset=ImageNet1k2022.0567.4