VECA

Vision Transformer uses core-periphery attention for linear scaling

Researchers have developed VECA, a novel Vision Transformer architecture that addresses the quadratic computational cost associated with high-resolution images. VECA utilizes an efficient linear-time attention mechanism…