Image Classification On Imagenet V2

Metrics

Top 1 Accuracy

Results

Performance results of various models on this benchmark

Model Name	Top 1 Accuracy	Paper Title
ResMLP-S24/16	69.8	ResMLP: Feedforward networks for image classification with data-efficient training
ResMLP-S12/16	66.0	ResMLP: Feedforward networks for image classification with data-efficient training
Mixer-B/8-SAM	65.5	When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
MAWS (ViT-6.5B)	84.0	The effectiveness of MAE pre-pretraining for billion-scale pretraining
LeViT-256	69.9	LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
CAIT-M36-448	76.7	Going deeper with Image Transformers
ViT-B-36x1	73.9	Three things everyone should know about Vision Transformers
MOAT-1 (IN-22K pretraining)	78.4	MOAT: Alternating Mobile Convolution and Attention Brings Strong Vision Models
ResNet50 (A1)	68.7	ResNet strikes back: An improved training procedure in timm
Discrete Adversarial Distillation (ViT-B, 224)	71.7	Distilling Out-of-Distribution Robustness from Vision-Language Foundation Models
SwinV2-B	78.08	Swin Transformer V2: Scaling Up Capacity and Resolution
Model soups (ViT-G/14)	84.22	Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
ViT-B/16-SAM	67.5	When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
LeViT-192	68.7	LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference
SEER (RegNet10B)	76.2	Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision
MOAT-2 (IN-22K pretraining)	79.3	MOAT: Alternating Mobile Convolution and Attention Brings Strong Vision Models
VOLO-D4	77.8	VOLO: Vision Outlooker for Visual Recognition
MOAT-3 (IN-22K pretraining)	80.6	MOAT: Alternating Mobile Convolution and Attention Brings Strong Vision Models
SwinV2-G	84.00%	Swin Transformer V2: Scaling Up Capacity and Resolution
ResMLP-B24/8	73.4	ResMLP: Feedforward networks for image classification with data-efficient training

0 of 33 row(s) selected.