List view

• microDL 2.1.0

  * Implement more diverse and configurable data augmentation, probably using [gunpowder](https://github.com/funkey/gunpowder). * Reduce data duplication - a clean data loader and augmenter will enable this.

  No due date

  •5/5 issues closed
  100% complete0 open5 closed

• microDL 2.2.0

  * Revisit config validation: the config validation can be implemented with modern libraries such as dataclasses/pydantic (https://pydantic-docs.helpmanual.io/usage/dataclasses/). * Revisit module structure and CLI: we can use Typer (https://typer.tiangolo.com/) or click (https://click.palletsprojects.com/en/8.1.x/) to write CLI.

  No due date
  0% complete0 open0 closed

• microDL 2.0.0

  This version of microDL will use pytorch for implementing 2D and 2.5D UNets for image translation. We will phase out 3D UNet and tensorflow code. We will leave the ability to train a segmentation model. Consolidate the parameters into one config, and select the stage of pipeline with a CLI (e.g., microDL --preprocess config.yml, microDL --train config.yml, or microDL --predict config.yml).

  No due date

  •15/16 issues closed
  93% complete1 open15 closed

• microDL 1.0.0

  We will start using release management to coordinate development of microDL.

  Overdue by 3 year(s)

  •

  Due by September 27, 2022

  •7/7 issues closed
  100% complete0 open7 closed