road map for june/july

[kaczmarj]

The plan is to check off all these boxes by the end of july...

• [x] datasets for classification / regression / autoencoding (#63)
• [ ] add model architectures
  • [ ] stylegan in keras
  • [x] variational inference models with tensorflow probability
  • [ ] progressive u-net (maybe a reach goal) (#64)
• [ ] share trained models in tf.keras format (savedmodel or hdf5)
  • [ ] stylegan for 2D MR slice generation (and maybe 3D)
  • [ ] bayesian model for brain mask
  • [x] bayesian model for freesurfer segmentation

This roadmap also gives us a chance to consider the scope of nobrainer. in my opinion, nobrainer should provide models, losses, metrics, layers, and TFRecord I/O for magnetic resonance imaging in 3D. also guides for how to use nobrainer. i think the training wrapper in nobrainer should be removed, and instead, examples should explain how to train different models (e.g., for semantic segmentation, GANs, classifiers, etc.).

cc: @wazeerzulfikar @satra

related to https://github.com/ohbm/hackathon2019/issues/79

[kaczmarj]

closed because this has gone stale