depth to space

A method that rearranges data from the depth dimension into spatial dimensions. It is often used in sub-pixel convolution layers of some super-resolution networks.