ldctbench.methods.wganvgg.network

Discriminator(input_size)

Bases: Module

Discriminator (Critic) for WGAN-VGG training

Parameters:

• input_size (int) –

  Input size of images fed in forward pass.

add_block(layers, ch_in, ch_out, stride) staticmethod

Append Conv -> LeakyReLU block to layer list.

Parameters:

• layers (List[Module]) –

  List of layers

• ch_in (int) –

  Number of input features

• ch_out (int) –

  Number of output features

• stride (int) –

  Desired stride of the conv layer

Returns:

• List[Module] –

  Layer list with appended layer.

conv_output_size(input_size, kernel_size_list, stride_list) staticmethod

Compute output size after feature extractor.

Parameters:

• input_size (int) –

  Input size of images fed in forward pass.

• kernel_size_list (List[int]) –

  List of kernel sizes for each layer.

• stride_list (List[int]) –

  List of strides for each layer.

Returns:

• int –

  Output size after feature extractor.

Model(args)

Bases: Module

Generator for WGAN-VGG

Parameters:

• args (Namespace) –

  Command line arguments passed to the model.