Domain Generalization on ImageNet-A

Metric: Top-1 accuracy % (higher is better)

LeaderboardDataset

Loading chart...

Results

Hide extra data

Sort:

#	Model↕	Top-1 accuracy %▼	Extra Data	Paper	Date↕	Code
1	Model soups (BASIC-L)	94.17	Yes	Model soups: averaging weights of multiple fine-...	2022-03-10	Code
2	Model soups (ViT-G/14)	92.67	Yes	Model soups: averaging weights of multiple fine-...	2022-03-10	Code
3	µ2Net+ (ViT-L/16)	84.53	Yes	A Continual Development Methodology for Large-sc...	2022-09-15	Code
4	CAR-FT (CLIP, ViT-L/14@336px)	81.5	Yes	Context-Aware Robust Fine-Tuning	2022-11-29	-
5	CAFormer-B36 (IN-21K, 384)	79.5	Yes	MetaFormer Baselines for Vision	2022-10-24	Code
6	MAE (ViT-H, 448)	76.7	No	Masked Autoencoders Are Scalable Vision Learners	2021-11-11	Code
7	FAN-Hybrid-L(IN-21K, 384)	74.5	Yes	Understanding The Robustness in Vision Transform...	2022-04-26	Code
8	ConvFormer-B36 (IN-21K, 384)	73.5	Yes	MetaFormer Baselines for Vision	2022-10-24	Code
9	CAFormer-B36 (IN-21K)	69.4	Yes	MetaFormer Baselines for Vision	2022-10-24	Code
10	ConvNeXt-XL (Im21k, 384)	69.3	Yes	A ConvNet for the 2020s	2022-01-10	Code
11	MAE+DAT (ViT-H)	68.92	No	Enhance the Visual Representation via Discrete A...	2022-09-16	Code
12	ConvFormer-B36 (IN-21K)	63.3	Yes	MetaFormer Baselines for Vision	2022-10-24	Code
13	Pyramid Adversarial Training Improves ViT (Im21k)	62.44	Yes	Pyramid Adversarial Training Improves ViT Perfor...	2021-11-30	Code
14	CAFormer-B36 (384)	61.9	No	MetaFormer Baselines for Vision	2022-10-24	Code
15	TransNeXt-Base (IN-1K supervised, 384)	61.6	No	TransNeXt: Robust Foveal Visual Perception for V...	2023-11-28	Code
16	TransNeXt-Small (IN-1K supervised, 384)	58.3	No	TransNeXt: Robust Foveal Visual Perception for V...	2023-11-28	Code
17	ConvFormer-B36 (384)	55.3	No	MetaFormer Baselines for Vision	2022-10-24	Code
18	SEER (RegNet10B)	52.7	Yes	Vision Models Are More Robust And Fair When Pret...	2022-02-16	Code
19	TransNeXt-Base (IN-1K supervised, 224)	50.6	No	TransNeXt: Robust Foveal Visual Perception for V...	2023-11-28	Code
20	CAFormer-B36	48.5	No	MetaFormer Baselines for Vision	2022-10-24	Code
21	TransNeXt-Small (IN-1K supervised, 224)	47.1	No	TransNeXt: Robust Foveal Visual Perception for V...	2023-11-28	Code
22	FAN-L-Hybrid+STL	46.1	No	Fully Attentional Networks with Self-emerging To...	2024-01-08	Code
23	ConvFormer-B36	40.1	No	MetaFormer Baselines for Vision	2022-10-24	Code
24	Pyramid Adversarial Training Improves ViT (384x384)	36.41	No	Pyramid Adversarial Training Improves ViT Perfor...	2021-11-30	Code
25	Sequencer2D-L	35.5	No	Sequencer: Deep LSTM for Image Classification	2022-05-04	Code
26	Discrete Adversarial Distillation (ViT-B/224)	31.8	No	Distilling Out-of-Distribution Robustness from V...	2023-11-02	Code
27	Diffusion Classifier	30.2	No	Your Diffusion Model is Secretly a Zero-Shot Cla...	2023-03-28	Code
28	RVT-B*	28.5	No	Towards Robust Vision Transformer	2021-05-17	Code
29	RVT-S*	25.7	No	Towards Robust Vision Transformer	2021-05-17	Code
30	RVT-Ti*	14.4	No	Towards Robust Vision Transformer	2021-05-17	Code
31	GFNet-S	14.3	No	Global Filter Networks for Image Classification	2021-07-01	Code
32	CutMix+MoEx (ResNet-50)	8.4	No	On Feature Normalization and Data Augmentation	2020-02-25	Code
33	Discrete Adversarial Distillation (ResNet-50)	7.7	No	Distilling Out-of-Distribution Robustness from V...	2023-11-02	Code
34	CutMix (ResNet-50)	7.3	No	CutMix: Regularization Strategy to Train Strong ...	2019-05-13	Code
35	Mixup (ResNet-50)	6.6	No	mixup: Beyond Empirical Risk Minimization	2017-10-25	Code
36	Cutout (ResNet-50)	4.4	No	Improved Regularization of Convolutional Neural ...	2017-08-15	Code
37	ResNet-50 (300 Epochs)	4.2	No	Deep Residual Learning for Image Recognition	2015-12-10	Code
38	Stylized ImageNet (ResNet-50)	2.3	Yes	ImageNet-trained CNNs are biased towards texture...	2018-11-29	Code
39	ResNet-50	0	No	-	-	Code

#1Model soups (BASIC-L)SOTA
94.17
Top-1 accuracy %· Extra Data· 2022-03-10
Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time Code
#2Model soups (ViT-G/14)
92.67
Top-1 accuracy %· Extra Data· 2022-03-10
Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time Code
#3µ2Net+ (ViT-L/16)
84.53
Top-1 accuracy %· Extra Data· 2022-09-15
A Continual Development Methodology for Large-scale Multitask Dynamic ML Systems Code
#4CAR-FT (CLIP, ViT-L/14@336px)
81.5
Top-1 accuracy %· Extra Data· 2022-11-29
Context-Aware Robust Fine-Tuning
#5CAFormer-B36 (IN-21K, 384)
79.5
Top-1 accuracy %· Extra Data· 2022-10-24
MetaFormer Baselines for Vision Code
#6MAE (ViT-H, 448)SOTA
76.7
Top-1 accuracy %· 2021-11-11
Masked Autoencoders Are Scalable Vision Learners Code
#7FAN-Hybrid-L(IN-21K, 384)
74.5
Top-1 accuracy %· Extra Data· 2022-04-26
Understanding The Robustness in Vision Transformers Code
#8ConvFormer-B36 (IN-21K, 384)
73.5
Top-1 accuracy %· Extra Data· 2022-10-24
MetaFormer Baselines for Vision Code
#9CAFormer-B36 (IN-21K)
69.4
Top-1 accuracy %· Extra Data· 2022-10-24
MetaFormer Baselines for Vision Code
#10ConvNeXt-XL (Im21k, 384)
69.3
Top-1 accuracy %· Extra Data· 2022-01-10
A ConvNet for the 2020s Code
#11MAE+DAT (ViT-H)
68.92
Top-1 accuracy %· 2022-09-16
Enhance the Visual Representation via Discrete Adversarial Training Code
#12ConvFormer-B36 (IN-21K)
63.3
Top-1 accuracy %· Extra Data· 2022-10-24
MetaFormer Baselines for Vision Code
#13Pyramid Adversarial Training Improves ViT (Im21k)
62.44
Top-1 accuracy %· Extra Data· 2021-11-30
Pyramid Adversarial Training Improves ViT Performance Code
#14CAFormer-B36 (384)
61.9
Top-1 accuracy %· 2022-10-24
MetaFormer Baselines for Vision Code
#15TransNeXt-Base (IN-1K supervised, 384)
61.6
Top-1 accuracy %· 2023-11-28
TransNeXt: Robust Foveal Visual Perception for Vision Transformers Code
#16TransNeXt-Small (IN-1K supervised, 384)
58.3
Top-1 accuracy %· 2023-11-28
TransNeXt: Robust Foveal Visual Perception for Vision Transformers Code
#17ConvFormer-B36 (384)
55.3
Top-1 accuracy %· 2022-10-24
MetaFormer Baselines for Vision Code
#18SEER (RegNet10B)
52.7
Top-1 accuracy %· Extra Data· 2022-02-16
Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision Code
#19TransNeXt-Base (IN-1K supervised, 224)
50.6
Top-1 accuracy %· 2023-11-28
TransNeXt: Robust Foveal Visual Perception for Vision Transformers Code
#20CAFormer-B36
48.5
Top-1 accuracy %· 2022-10-24
MetaFormer Baselines for Vision Code
#21TransNeXt-Small (IN-1K supervised, 224)
47.1
Top-1 accuracy %· 2023-11-28
TransNeXt: Robust Foveal Visual Perception for Vision Transformers Code
#22FAN-L-Hybrid+STL
46.1
Top-1 accuracy %· 2024-01-08
Fully Attentional Networks with Self-emerging Token Labeling Code
#23ConvFormer-B36
40.1
Top-1 accuracy %· 2022-10-24
MetaFormer Baselines for Vision Code
#24Pyramid Adversarial Training Improves ViT (384x384)
36.41
Top-1 accuracy %· 2021-11-30
Pyramid Adversarial Training Improves ViT Performance Code
#25Sequencer2D-L
35.5
Top-1 accuracy %· 2022-05-04
Sequencer: Deep LSTM for Image Classification Code
#26Discrete Adversarial Distillation (ViT-B/224)
31.8
Top-1 accuracy %· 2023-11-02
Distilling Out-of-Distribution Robustness from Vision-Language Foundation Models Code
#27Diffusion Classifier
30.2
Top-1 accuracy %· 2023-03-28
Your Diffusion Model is Secretly a Zero-Shot Classifier Code
#28RVT-B*SOTA
28.5
Top-1 accuracy %· 2021-05-17
Towards Robust Vision Transformer Code
#29RVT-S*
25.7
Top-1 accuracy %· 2021-05-17
Towards Robust Vision Transformer Code
#30RVT-Ti*
14.4
Top-1 accuracy %· 2021-05-17
Towards Robust Vision Transformer Code
#31GFNet-S
14.3
Top-1 accuracy %· 2021-07-01
Global Filter Networks for Image Classification Code
#32CutMix+MoEx (ResNet-50)SOTA
8.4
Top-1 accuracy %· 2020-02-25
On Feature Normalization and Data Augmentation Code
#33Discrete Adversarial Distillation (ResNet-50)
7.7
Top-1 accuracy %· 2023-11-02
Distilling Out-of-Distribution Robustness from Vision-Language Foundation Models Code
#34CutMix (ResNet-50)SOTA
7.3
Top-1 accuracy %· 2019-05-13
CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features Code
#35Mixup (ResNet-50)SOTA
6.6
Top-1 accuracy %· 2017-10-25
mixup: Beyond Empirical Risk Minimization Code
#36Cutout (ResNet-50)SOTA
4.4
Top-1 accuracy %· 2017-08-15
Improved Regularization of Convolutional Neural Networks with Cutout Code
#37ResNet-50 (300 Epochs)SOTA
4.2
Top-1 accuracy %· 2015-12-10
Deep Residual Learning for Image Recognition Code
#38Stylized ImageNet (ResNet-50)
2.3
Top-1 accuracy %· Extra Data· 2018-11-29
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness Code
#39ResNet-50
0
Top-1 accuracy %
No paperCode