Release candidate 3 of v4

This is a develop release of v4.
The stable v4 will be released after the paper is drafted.

What's Changed

• [feat] itk_combine to combine multiple label sources by @MGAMZ in #61
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #62
• Support Runtime Override InferenceConfig by @MGAMZ in #63
• SegFormer3D now supports any stages by @MGAMZ in #65
• Add 3D Slicer extension for ITKIT inference by @Copilot in #64
• Align GUI tabs with backend CLI parameters by @Copilot in #66
• Add itk_slicer CLI entry point for SlicerITKIT server by @Copilot in #67
• Example, Log and Doc refine by @MGAMZ in #68
• Potential fix for code scanning alert no. 1: Workflow does not contain permissions by @MGAMZ in #69
• Release Docker Images by @MGAMZ in #70

Full Changelog: v4.0.0rc2...v4.0.0rc3

Release Candidate 2 of v4

This is a develop release of v4.
The stable v4 will be released after the paper is drafted.

What's Changed

• Prepare Release V4 by @MGAMZ in #49
• Compatible with mmengine loop arg by @MGAMZ in #50
• Fix metadata loss when processing modules skip existing files by @Copilot in #51
• Improve readthedoc layout structure by @MGAMZ in #52
• Introduce codecov into GitHub Action by @MGAMZ in #53
• Refactor test files structure and improve codecov by @MGAMZ in #54
• Introduce pre-commit ci APP by @MGAMZ in #55
• Support ONNX backend for itkit.mm.inference by @MGAMZ in #56
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #57
• Better InferenceConfig and onnx infer backend by @MGAMZ in #58
• Support MSD dataset by @MGAMZ in #59
• yank back version by @MGAMZ in #60

New Contributors

• @pre-commit-ci[bot] made their first contribution in #57

Full Changelog: v4.0.0rc1...v4.0.0rc2

Release Candidate of v4

ITKIT is experimencing BC before formally published. So v3.6 will not be published as a stable release.

We are entering v4 epoch and resume from v3.6.0rc0.

What's Changed

• BC for Introducing ITKIT v4 by @MGAMZ in #45
• Lint Related Improvements by @MGAMZ in #48

Full Changelog: v3.6.0rc0...v4.0.0rc1

Release Candidate of v3.6

What's Changed

• Purge SegFormer3D implementation and fix itk_patch tests by @MGAMZ in #30
• [Feature] itk_evaluate, itk_convert, torchio and more by @MGAMZ in #36
• Introduce MONAI-Based Train-Time Patching by @MGAMZ in #37
• Add IO shape validation tests for neural network models by @Copilot in #38
• Add comprehensive unit tests for itkit.io module by @Copilot in #40
• Add dataset mode to itk_orient for automatic image/label processing by @Copilot in #41
• Official example test on SegFormer3D, MONAI, TorchIO by @MGAMZ in #42
• Fix Inconsistent logic between TorchIO and ITKIT data backend by @MGAMZ in #43
• Compensate the missing commits in #43 by @MGAMZ in #44

Full Changelog: v3.5.3...v3.6.0rc0

Release v3.5.3

What's Changed

• Lint itkit.mm folder and deploy pyright into pre-commit by @MGAMZ in #25
• Lint more modules by @MGAMZ in #26
• Fix InferenceConfig not effecting during inference by @MGAMZ in #27
• Consistently orient to LPI during itk_patch by @MGAMZ in #28
• Fix IoUMetric bug by @MGAMZ in #29

Full Changelog: v3.5.0...v3.5.3

Release v3.5

Highlights

• Introducing pre-commit into this repo.
• Quality improvements by: pyupgrade, ruff, isort.
• Python static check by: pyright, mypy.
• Accelerated PatchDataset indexing during the initialization of training.

PRs

• Add ITKIT LOGO by @MGAMZ in #21
• Introduce auto lint and static check by @MGAMZ in #22
• Accelerate Indexing of PatchDataset by @MGAMZ in #23

Full Changelog: v3.4.0...v3.5.0

Release v3.4

Full Changelog: v3.3.0...v3.4.0

Highlights

1. In this version, I refactor itk_processing modules to provide better maintainability and code structure.
2. CICD has also undergone improvements, resulting in higher module stability for this library.
3. The mmengine based sliding-window inference efficiency has been improved via introducing a patch cache which can accelerate Device2Host transfer.

New Features

• Support Totalsegmentator to CTSpine1K label extraction
• Support Totalsegmentator patched dataset
• An evaluation script for 3D medical volumes with segmentation tasks.

Enhancements

• GitHub Action environments upgrade:
  • actions/checkout
  • actions/setup-python
• Push will only trigger CI check with python=3.13.
• Disable pin_memory for mgam_model_Seg3D inference cache pred, count_mat to save peak RAM occupation.
• Introduce pinned patch cache on host to accelerate Device2Host transfer, this can make sliding-window inference much faster.
• Refactor itk_process modules to provide better readability and maintainance experiments.
• Add pytest units for itk_process modules.
• Auto publish to TestPyPI when doing a pre-release.
• Other general lint fixes.

Bug Fixes

• itk_process module produces incorrect meta json file containing only one series.
• itk_resample normalize files' suffix differently in two input modes.

Repo Clean

• Remove unused packages, including precrop_npz, standard_dataset, Sup_2D.
• Remove stretch processing module, it's one of my previous standalone work.
• Remove unused sitk_io functions.
• Remove inferencer of 2D mode, ONNX mode.
• Now using pyright to check this repo, there're still many errors require fixing.
• Remove statistics, search_tool, SlideAvg, ckptclean, OpticalVisualization, ConfigQuickModify, CalcLabelDifference as they are strongly relied on specific dataset or task storage structure.
• Remove ResampleMha script as itk_resample is a better one with CLI.

Related PRs

• Upgrade 3.4 by @MGAMZ in #14
• Auto publish to TestPyPI when doing a pre-release by @MGAMZ in #15
• Upgrade Release Candidate of v3.4 by @MGAMZ in #17
• Bump v3.4.0rc2 by @MGAMZ in #18
• Bump v3.4.0rc3 by @MGAMZ in #19

Release Candidate 3 of v3.4

Bump v3.4.0rc3 by @MGAMZ in #19

Modification

1. Delete unused srcs, test units, package imports
2. Introduce evaluator util.
3. Remove deprecated codes in Inferencer.
4. Private methods are behind the public methods from now on, which will act as a coding guideline.
5. Bug Fixes

Full Changelog: v3.4.0rc2...v3.4.0rc3

Release Candidate 2 of v3.4

What's Changed between rc1 and rc2

I'm trying to integrate the linter as pre-commit or CI. But no one satisfies me.

1. Fix several lint suggestions. There're too many lint errors, it's hard to solve them in short time, so leaving them ready to be solved. pyright has been selected to perform static type check.
2. Fix the incorrect arg passing in SingleResampleProcessor.
3. Add patch_size assertion check to avoid generating patches with wrong sizes. They can cause problems during deep learning training phase in which is quite complicated to locate the root of this issue.
4. Decouple the IoU_metric logic from mmseg to ITKIT, which perform evaluation on Host.
5. Change optional dependency.
6. All pre_crop scripts are removed as the new itk_patch is recommended.
7. Improve several test unit design.

Full Changelog: v3.4.0rc1...v3.4.0rc2

Release Candidate 1 of v3.4

What's Changed

• Upgrade Release Candidate of v3.4 by @MGAMZ in #17

Full Changelog: v3.4.0rc0...v3.4.0rc1