niftynet.layer.elementwise module

class ElementwiseLayer(func, initializer=None, regularizer=None, name='residual')[source]

Bases: niftynet.layer.base_layer.TrainableLayer

This class takes care of the elementwise sum in a residual connection It matches the channel dims from two branch flows, by either padding or projection if necessary.

layer_op(param_flow, bypass_flow)[source]