Domain Generalization On Imagenet R

Metrics

Top-1 Error Rate

Results

Performance results of various models on this benchmark

Model Name	Top-1 Error Rate	Paper Title	Repository
Model soups (ViT-G/14)	4.54	Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time	-
GPaCo (ViT-L)	39.7	Generalized Parametric Contrastive Learning	-
SEER (RegNet10B)	43.9	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision	-
Pyramid Adversarial Training Improves ViT (Im21k)	42.16	Pyramid Adversarial Training Improves ViT Performance	-
VOLO-D5+HAT	40.3	Improving Vision Transformers by Revisiting High-frequency Components	-
Stylized ImageNet (ResNet-50)	58.5	ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness	-
Mixer-B/8-SAM	76.5	When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations	-
LLE (ViT-B/16, SWAG, Edge Aug)	31.3	A Whac-A-Mole Dilemma: Shortcuts Come in Multiples Where Mitigating One Amplifies Others	-
Model soups (BASIC-L)	3.90	Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time	-
MAE (ViT-H, 448)	33.5	Masked Autoencoders Are Scalable Vision Learners	-
PRIME with JSD (ResNet-50)	53.7	PRIME: A few primitives can boost robustness to common corruptions	-
ConvFormer-B36	48.9	MetaFormer Baselines for Vision	-
CAR-FT (CLIP, ViT-L/14@336px)	10.3	Context-Aware Robust Fine-Tuning	-
ConvFormer-B36 (384)	47.8	MetaFormer Baselines for Vision	-
ConvNeXt-XL (Im21k, 384)	31.8	A ConvNet for the 2020s	-
ResNet-152x2-SAM	71.9	When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations	-
FAN-Hybrid-L(IN-21K, 384))	28.9	Understanding The Robustness in Vision Transformers	-
RVT-Ti*	56.1	Towards Robust Vision Transformer	-
RVT-S*	52.3	Towards Robust Vision Transformer	-
ConvFormer-B36 (IN21K, 384)	33.5	MetaFormer Baselines for Vision	-

0 of 39 row(s) selected.