Skip to content

ldctbench.methods.dugan.network

UNet(repeat_num, use_tanh=False, use_sigmoid=False, skip_connection=True, use_discriminator=True, conv_dim=64, in_channels=1)

Bases: Module

Parameters:

  • repeat_num (int) –

    Number of downsampling steps

  • use_tanh (bool, default: False ) –

    Use tanh as final nonlinearity, by default False

  • use_sigmoid (bool, default: False ) –

    Use sigmoid as final nonlinearity, by default False

  • skip_connection (bool, default: True ) –

    Use skip connection, by default True

  • use_discriminator (bool, default: True ) –

    Use encoder (at bottleneck) as additional output, by default True

  • conv_dim (int, default: 64 ) –

    Number of features in first layer, by default 64

  • in_channels (int, default: 1 ) –

    Number of channels of first layer, by default 1