Revisiting Weakly Supervised Pre-Training of Visual Perception Models

Open paper
Benchmark Model Rank Results
fine-grained-image-classification-on-cub-200-1SWAG (ViT H/14)#5Accuracy: 91.7
image-classification-on-imagenetSWAG (ViT H/14)#36Top 1 Accuracy: 88.6%Number of params: 633.5MGFLOPs: 1018.8
image-classification-on-imagenet-realSWAG (RegNetY 128GF)#13Accuracy: 90.7%
image-classification-on-imagenet-v2SWAG (ViT H/14)#9Top 1 Accuracy: 81.1
image-classification-on-inaturalist-2018SWAG (ViT H/14)#6Top-1 Accuracy: 86.0%
image-classification-on-objectnetSWAG (ViT H/14)#13Top-1 Accuracy: 69.5
image-classification-on-objectnetRegNetY 128GF (Platt)#15Top-1 Accuracy: 64.3
image-classification-on-objectnetViT H/14 (Platt)#18Top-1 Accuracy: 60
image-classification-on-objectnetViT L/16 (Platt)#20Top-1 Accuracy: 57.3
image-classification-on-objectnetViT B/16#25Top-1 Accuracy: 48.9
image-classification-on-places365-standardSWAG (ViT H/14)#1Top 1 Accuracy: 60.7

Rank counts only results with a code link.