ldctbench.methods.resnet.network

Model(args, n_channels=128, n_blocks=10)

Bases: Module

Residual network

As proposed in the paper, the network performs noise subtraction on the input.

Parameters:

• args (Namespace) –

  Command line arguments passed to the model.

• n_channels (int, default: 128 ) –

  Number of features for each conv layer, by default 128

• n_blocks (int, default: 10 ) –

  Number of residual blocks, by default 10

ResBlock(ch)

Bases: Module

Single Residual block

Each block consists of Conv->BN->ReLU->GroupConv->BN->ReLU->Conv + |__________|

Parameters:

• ch (int) –

  Number of features to use for each conv layer.