project-monai / monai Goto Github PK

-Tensorboard visualisation of 3d images

a subtask of #78, port dense v-net

ref:
https://niftynet.readthedocs.io/en/dev/niftynet.network.dense_vnet.html

@eligibson

Sliding window inference
Writing segmentation as a nifti image (Nifti image writer)
Writing classification as a json file (json writer)

-Refactor input ‘transform’ interfaces
- Finalise the code folder structure
- More thorough unit tests
- More integration tests
- Following the current design, more networks/losses/metrics/transforms

Definition of Networks, standardized way of defining network parameters

e.g. Unzip .tar; read .json
definition and create a standardized way of constructing input data pipelines

a subtask of #78, port highresnet

ref:
https://niftynet.readthedocs.io/en/dev/niftynet.network.highres3dnet.html
https://github.com/fepegar/highresnet

-Depends on task 2

• Minimal evaluation metrics interface (or using ignite.metrics?)
  -Event handler for model validation and early stopping

(This ticket is originally from the monai-minimal repo)

Would be great to have a good coding style from day one...
e.g.,
Function and variable names should be lowercase, with words separated by underscores as necessary to improve readability.
Class names should normally use the CapWords convention.

To close this ticket, we ensure that on the project running flake8 (with pep8-naming plugin) with config https://github.com/Project-MONAI/MONAI/blob/a106c79b706a72933520aaba4fc43eb85765985e/.flake8
returns no error message.

Additional comment from #21

We should also nail down what format comments should be in, I would suggest Numpy format https://numpydoc.readthedocs.io/en/latest/format.html as a good hybrid of Google docstrings and reStructured Text formats, however Pytorch uses Google docstrings so maybe we should stick to that.

-Additional refactoring iteration to unify single-gpu/multiple gpu training?