Learned Queries for Efficient Local Attention

Open paper
Benchmark Model Rank Results
image-classification-on-imagenetQnA-ViT-Base#382Top 1 Accuracy: 83.7%Number of params: 56MGFLOPs: 9.7
image-classification-on-imagenetQnA-ViT-Small#442Top 1 Accuracy: 83.2%Number of params: 25MGFLOPs: 4.4
image-classification-on-imagenetQnA-ViT-Tiny#595Top 1 Accuracy: 81.7%Number of params: 16MGFLOPs: 2.5

Rank counts only results with a code link.