Skip to content

Releases: Douwe-Spaanderman/InteractiveNet

v0.2.2

Choose a tag to compare

@Douwe-Spaanderman Douwe-Spaanderman released this 13 Feb 10:01

Change Log

0.2.2

Released on February 13, 2024.

Fixed

  • Correct reference in README

Added

  • New faster preprocessing method for GUI. Previosly resampling was done prior to cropping. Now first a large cropping is done, followed by resampling and finally a tighter crop is done. This drastically speed up pre- and post-processing, especially for larger volumetric images with small objects of interest.

v0.2.1

Choose a tag to compare

@Douwe-Spaanderman Douwe-Spaanderman released this 26 Oct 10:46

Change Log

0.2.1

Released on October 26, 2023.

Fixed

  • Improve Docker image
  • Fixed bug when no test segmentations were present in generating dataset json
  • Fixed bug where self-trained model was loaded while using deployed model

Added

  • Support for MONAI 1.2

v0.2.0

Choose a tag to compare

@Douwe-Spaanderman Douwe-Spaanderman released this 20 Jun 12:31
f0662e6

Change Log

0.2.0

Released on June 20, 2023.

Added

  • Zenodo release
  • Release on Pypi
  • LICENSE added to all files
  • Documentation for releases
  • WORC models available on https://doi.org/10.5281/zenodo.8054038
  • Models now downloadable and installable
  • Added docker image for GUI (or atleast monailabel server for GUI)

Changed

  • Personal author to group (BIGR)

v0.1.0

Choose a tag to compare

@Douwe-Spaanderman Douwe-Spaanderman released this 14 Jun 13:33

Change Log

0.1.0

Released on June 14, 2023.

Added

  • First version of the Minimally Interactive Segmentation Method, containing:
    • InteractiveNet: a self-configuring workflow for finding the best configuration for preprocessing, training, postprocessing, ensembling based on dataset characteristics for interactive segmentation.
    • Generating synthetic interactions from ground truth segmentation, in order to pilot interactive segmentation.
    • A graphical user interface (GUI) in order to use InteractiveNet in 3D slicer or OHIF
    • Deployment of trained models
    • Licence, documentation and setup files.