Folding Architectures Cut Hardware and Power Overhead in Two-Dimensional Image Convolution
Researchers have introduced folded separable convolution architectures that restrict hardware requirements to two multipliers and two adders regardless of kernel ...


