62 citations · 97 across the 17 of their papers we have counts for
1 paper · 1 filter
Shunya Shimomura, Kazuhiro Hotta
Vision Transformer (ViT) has been widely used as a powerful framework for modeling global dependencies among image patches. However, its core component, self-attention assigns soft…