mridc.collections.reconstruction.models.conv package

Submodules

mridc.collections.reconstruction.models.conv.conv2d module

class mridc.collections.reconstruction.models.conv.conv2d.Conv2d(in_channels, out_channels, hidden_channels, n_convs=3, activation=PReLU(num_parameters=1), batchnorm=False)[source]

Bases: Module

Implementation of a simple cascade of 2D convolutions. If batchnorm is set to True, batch normalization layer is applied after each convolution.

forward(x)[source]

Performs the forward pass of Conv2d.

Parameters

x (Input tensor.) –

Return type

Convoluted output.

training: bool

mridc.collections.reconstruction.models.conv.gruconv2d module

class mridc.collections.reconstruction.models.conv.gruconv2d.GRUConv2d(in_channels, out_channels, hidden_channels, n_convs=3, activation='ReLU', batchnorm=False)[source]

Bases: Module

Implementation of a GRU followed by a number of 2D convolutions inspired by 1.

References

1

  1. Qin, J. Schlemper, J. Caballero, A. N. Price, J. V. Hajnal and D. Rueckert, “Convolutional Recurrent Neural Networks for Dynamic MR Image Reconstruction,” in IEEE Transactions on Medical Imaging, vol. 38, no. 1, pp. 280-290, Jan. 2019, doi: 10.1109/TMI.2018.2863670.

forward(x, hx: Optional[Tensor] = None)[source]

Performs the forward pass of Conv2d.

Parameters

  • x (Input tensor.) – torch.Tensor

  • hx (Initial hidden state.) – torch.Tensor

Return type

Convoluted output.

training: bool

Module contents