List your company
ResearchFormerly WizWand

Image Classification on ImageNet v2 56 (test)

84AccuracySwinV2-G
80.25681.22882.283.172May 22, 2023
Updated 8mo ago

Evaluation results

MethodLinks
SwinV2-GPretraining=IN-21k + 70M, Input resolution=640x640, Parameters=3000M, Fine-tuning=true2023.0584
SoViT-400m/14Pretraining=JFT-3B, Input resolution=518x518, Parameters=428M, FLOPs=1374G, Fine-tuning=true2023.0583.4
ViT-G/14Pretraining=JFT-3B, Input resolution=518x518, Parameters=1882M, FLOPs=5668G, Fine-tuning=true2023.0583.3
SoViT-400m/14Pretraining=JFT-3B, Input resolution=384x384, Parameters=428M, FLOPs=672G, Fine-tuning=true2023.0583.2
MAE-WSPPretraining=IG-3B, Input resolution=518x518, Parameters=1890M, FLOPs=5679G, Fine-tuning=true2023.0583
SoViT-400m/14Pretraining=JFT-3B, Input resolution=224x224, Parameters=428M, FLOPs=221G, Fine-tuning=true2023.0580.7
ViT-L/16Pretraining=JFT-3B, Input resolution=384x384, Parameters=303M, FLOPs=383G, Fine-tuning=true2023.0580.4