/Segm_Net_3D

This library is an implementation of a 3-D U-net using dice loss and batch normalization, implemented in TensorFlow.

Primary LanguageJupyter NotebookMIT LicenseMIT

TensorFlow 3-D Segmentation Networks Library

This library contains at the moment:

  • 3-D U-net (or V-net) using dice loss and batch normalization
  • 3-D Convolutional Autoencoder using dice loss/crossentropy and batch normalization

How do I set it up?

  • TensorFlow 1.12 (GPU recomended)
  • Configuration: Just import it
  • Dependencies: Numpy, skimage (for shape generation in example)
  • How to run tests: Open the jupyter notebooks and hit "run"