1 paper
Zhe Feng, Longfei Liu, Wei Liu +7
Self-attention-based vision backbones perform well on dense prediction, but the quadratic computational cost of global softmax attention limits their efficiency as image resolution…