UpSample2x#

tiatoolbox . models . architecture . utils . UpSample2x

class UpSample2x[source]#

A layer to scale input by a factor of 2.

This layer uses Kronecker product underneath rather than the default pytorch interpolation.

Methods

Logic for using layers defined in init.

Attributes

forward(x)[source]#

Logic for using layers defined in init.

Parameters:: x (torch.Tensor) – Input images, the tensor is in the shape of NCHW.
Returns:: Input images upsampled by a factor of 2 via nearest neighbour interpolation. The tensor is the shape as NCHW.
Return type:: torch.Tensor