Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations

Open paper
Benchmark Model Rank Results
efficient-vits-on-imagenet-1k-with-deit-sEViT (80%)#8Top 1 Accuracy: 79.8GFLOPs: 3.5
efficient-vits-on-imagenet-1k-with-deit-sEViT (90%)#9Top 1 Accuracy: 79.8GFLOPs: 4.0
efficient-vits-on-imagenet-1k-with-deit-sEViT (70%)#19Top 1 Accuracy: 79.5GFLOPs: 3.0
efficient-vits-on-imagenet-1k-with-deit-sEViT (60%)#29Top 1 Accuracy: 78.9GFLOPs: 2.6
efficient-vits-on-imagenet-1k-with-deit-sEViT (50%)#34Top 1 Accuracy: 78.5GFLOPs: 2.3
efficient-vits-on-imagenet-1k-with-lv-vit-sEViT (70%)#12Top 1 Accuracy: 83.0GFLOPs: 4.7
efficient-vits-on-imagenet-1k-with-lv-vit-sEViT (50%)#16Top 1 Accuracy: 82.5GFLOPs: 3.9

Rank counts only results with a code link.