Skip to content

Latest commit

 

History

History
56 lines (42 loc) · 1.94 KB

README.md

File metadata and controls

56 lines (42 loc) · 1.94 KB
Raw

NN Common Modules

Common modules, blocks and losses which can be reused in a deep neural netwok specifically for segmentation Please use technical documentation for a reference to API manual

This project has 3 modules

  • Losses (losses.py) -> It has all the loss functions defined as python classes

    1. DiceLoss
    2. IoULoss
    3. CrossEntropyLoss2d
    4. CombinedLoss
    5. FocalLoss

  • Modules (modules.py) -> It has all the commonly used building blocks of an FCN

    1. DenseBlock
    2. EncoderBlock
    3. DecoderBlock
    4. ClassifierBlock
    5. GenericBlock
    6. SDNetEncoderBlock
    7. SDNetDecoderBlock
    8. OctaveDenseBlock
    9. OctaveEncoderBlock
    10. OctaveDecoderBlock

  • Bayesian Modules (bayesian_modules.py) -> It has all the commonly used building blocks of a Bayesian FCN

    1. BayesianConvolutionBlock
    2. BayesianEncoderBlock
    3. BayesianDecoderBlock
    4. BayesianClassifierBlock

Getting Started

Pre-requisites

You need to have following in order for this library to work as expected

  1. Python >= 3.5
  2. Pytorch >= 1.0.0
  3. Numpy >= 1.14.0

Installing

Always use the latest release. Use following command with appropriate version no(v1.2) in this particular case to install. You can find the link for the latest release in the release section of this github repo

pip install https://github.com/ai-med/nn-common-modules/releases/download/v1.2/nn_common_modules-1.4-py3-none-any.whl

Authors

Help us improve

Let us know if you face any issues. You are always welcome to report new issues and bugs and also suggest further improvements. And if you like our work hit that start button on top. Enjoy :)