Domain Generalization On Imagenet C

Metrics

Top 1 Accuracy

Results

Performance results of various models on this benchmark

Model Name	Top 1 Accuracy	Paper Title	Repository
Mixer-B/8-SAM	48.9	When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations	-
APR-SP + DeepAugment (ResNet-50)	-	Amplitude-Phase Recombination: Rethinking Robustness of Convolutional Neural Networks in Frequency Domain	-
CAFormer-B36 (IN21K, 384)	-	MetaFormer Baselines for Vision	-
CAFormer-B36 (IN21K)	-	MetaFormer Baselines for Vision	-
FAN-L-Hybrid	67.7	Understanding The Robustness in Vision Transformers	-
GFNet-S	-	Global Filter Networks for Image Classification	-
MAE (ViT-H)	-	Masked Autoencoders Are Scalable Vision Learners	-
ResNet-50 (PushPull-Conv) + PRIME	69.4	PushPull-Net: Inhibition-driven ResNet robust to image corruptions	-
ConvNeXt-XL (Im21k) (augmentation overlap with ImageNet-C)	-	A ConvNet for the 2020s	-
FAN-L-Hybrid+STL	69.2	Fully Attentional Networks with Self-emerging Token Labeling	-
ConvFormer-B36	-	MetaFormer Baselines for Vision	-
QualNet (ResNeXt101)	-	Quality-Agnostic Image Recognition via Invertible Decoder
APR-SP (ResNet-50)	-	Amplitude-Phase Recombination: Rethinking Robustness of Convolutional Neural Networks in Frequency Domain	-
DINOv2 (ViT-S/14, frozen model, linear eval)	-	DINOv2: Learning Robust Visual Features without Supervision	-
DiscreteViT	-	Discrete Representations Strengthen Vision Transformer Robustness	-
DiffAUD (Swin-Tiny)	61	Diffusion-Based Adaptation for Classification of Unknown Degraded Images
AugMix (ResNet-50)	-	AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty	-
DINOv2 (ViT-g/14, frozen model, linear eval)	-	DINOv2: Learning Robust Visual Features without Supervision	-
DiffAUD (ResNet-50)	52.1	Diffusion-Based Adaptation for Classification of Unknown Degraded Images
GPaCo (ViT-L)	-	Generalized Parametric Contrastive Learning	-

0 of 47 row(s) selected.