tiramisu_brulee.model package

Submodules

tiramisu_brulee.model.dense module

Blocks/layers for densely-connected networks Author: Jacob Reinhold <jcreinhold@gmail.com> Created on: Jul 02, 2020

class tiramisu_brulee.model.dense.Bottleneck2d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.Bottleneck

class tiramisu_brulee.model.dense.Bottleneck3d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.Bottleneck

class tiramisu_brulee.model.dense.DenseBlock2d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.DenseBlock

class tiramisu_brulee.model.dense.DenseBlock3d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.DenseBlock

class tiramisu_brulee.model.dense.TransitionDown2d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.ConvLayer

class tiramisu_brulee.model.dense.TransitionDown3d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.ConvLayer

class tiramisu_brulee.model.dense.TransitionUp2d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.TransitionUp

class tiramisu_brulee.model.dense.TransitionUp3d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.dense.TransitionUp

tiramisu_brulee.model.tiramisu module

PyTorch Tiramisu network

PyTorch implementation of the Tiramisu network architecture. Implementation based on pytorch_tiramisu.

Changes from pytorch_tiramisu include:

  1. removal of bias from conv layers,

  2. change zero padding to replication padding,

  3. cosmetic changes for brevity, clarity, consistency

References

Jégou, Simon, et al. “The one hundred layers tiramisu: Fully convolutional densenets for semantic segmentation.” CVPR. 2017.

https://github.com/bfortuner/pytorch_tiramisu

Author: Jacob Reinhold <jcreinhold@gmail.com> Created on: Jul 01, 2020

class tiramisu_brulee.model.tiramisu.Tiramisu2d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.tiramisu.Tiramisu

class tiramisu_brulee.model.tiramisu.Tiramisu3d(*args: Any, **kwargs: Any)[source]

Bases: tiramisu_brulee.model.tiramisu.Tiramisu

Module contents