niftynet.network.se_resnet module¶

class SE_ResNetDesc(bn, fc, conv1, blocks)¶

Bases: tuple

blocks¶: Alias for field number 3

bn¶: Alias for field number 0

conv1¶: Alias for field number 2

fc¶: Alias for field number 1

class SE_ResNet(num_classes, n_features=[16, 64, 128], n_blocks_per_resolution=1, w_initializer=None, w_regularizer=None, b_initializer=None, b_regularizer=None, acti_func='relu', name='SE_ResNet')[source]¶

Bases: niftynet.network.base_net.BaseNet

3D implementation of SE-ResNet:: Hu et al., “Squeeze-and-Excitation Networks”, arXiv:1709.01507v2

create()[source]¶

layer_op(images, is_training=True, **unused_kwargs)[source]¶

class BottleneckBlockDesc1(conv)¶

Bases: tuple

conv¶: Alias for field number 0

class BottleneckBlockDesc2(common_bn, conv, conv_shortcut)¶

Bases: tuple

common_bn¶: Alias for field number 0

conv¶: Alias for field number 1

conv_shortcut¶: Alias for field number 2

class BottleneckBlock(n_output_chns, stride, Conv, name='bottleneck')[source]¶

Bases: niftynet.layer.base_layer.TrainableLayer

create(input_chns)[source]¶

layer_op(images, is_training=True)[source]¶

class DownResBlockDesc(blocks)¶

Bases: tuple

blocks¶: Alias for field number 0

class DownResBlock(n_output_chns, count, stride, Conv, name='downres')[source]¶

Bases: niftynet.layer.base_layer.TrainableLayer

create()[source]¶

layer_op(images, is_training)[source]¶