dltk.models.segmentation package

dltk.models.segmentation.fcn module

class dltk.models.segmentation.fcn.ResNetFCN(num_classes, num_residual_units=3, filters=(16, 64, 128, 256, 512), strides=((1, 1, 1), (2, 2, 2), (2, 2, 2), (2, 2, 2), (1, 1, 1)), relu_leakiness=0.1, name='resnetfcn')

Bases: dltk.core.modules.base.AbstractModule

FCN module with residual encoder

This module builds a FCN for segmentation using a residual encoder.

class dltk.models.segmentation.fcn.Upscore(out_filters, strides, name='upscore')

Bases: dltk.core.modules.base.AbstractModule

Upscore module according to J. Long.

dltk.models.segmentation.unet module

class dltk.models.segmentation.unet.ResUNET(num_classes, num_residual_units=3, filters=(16, 64, 128, 256, 512), strides=((1, 1, 1), (2, 2, 2), (2, 2, 2), (2, 2, 2), (1, 1, 1)), relu_leakiness=0.1, name='resnetfcn')

Bases: dltk.core.modules.base.SaveableModule

ResUNET module with residual encoder

This module builds a UNET for segmentation using a residual encoder.

output_keys = ['logits', 'y_prob', 'y_']

class dltk.models.segmentation.unet.UpsampleAndConcat(strides, name='upandconcat')

Bases: dltk.core.modules.base.AbstractModule

UNET upsampling module according to O. Ronneberger.

Module contents