Version 0.6.1

Wsinfer, now with strided patches.

This is a minor release that adds strided patches, via the --patch-overlap-ratio option to wsinfer run.

This release also includes various bugfixes and a move to the ruff formatter/linter.

What's Changed

• Cleanup + add strided patches by @kaczmarj in #218
• [STYLE] use ruff in lieu of black/flake8/isort + add pre-commit by @kaczmarj in #219

Full Changelog: v0.6.0...v0.6.1

Version 0.6.0

Welcome, WSInfer version 0.6.0! This version makes the tissue segmentation step more customizable, with options beginning with --seg- to wsinfer run. Python 3.12 is also supported now, and a QuPath project can be created with WSInfer results using the --qupath option in wsinfer run. See below for other changes.

What's Changed

• [DOCS] add video walkthrough to user guide + update version switcher by @kaczmarj in #201
• Fix/recursion limit by @swaradgat19 in #208
• fix types when using cv2 by @kaczmarj in #211
• Feat/qupath by @swaradgat19 in #209
• [STYLE] format with black by @kaczmarj in #215
• add python 3.12 to ci by @kaczmarj in #197
• Add segmentation options to wsinfer run by @kaczmarj in #217

Full Changelog: v0.5.0...v0.6.0

Version 0.5.0

This is a minor version of WSInfer. One change is that GeoJSON files of model outputs are created automatically after inference. See other fixes below.

What's Changed

• check that page0 is TiffPage inst by @kaczmarj in #188
• restrict wsi_dir.glob to files by @xalexalex in #190
• Fix/geojson command by @swaradgat19 in #191
• test that openslide.OpenSlide is available by @kaczmarj in #199
• clarify docs on gpu install + explain model output directory by @kaczmarj in #200

New Contributors

• @xalexalex made their first contribution in #190

Full Changelog: v0.4.1...v0.5.0

Version 0.4.1

What's Changed

• remove vgg16 and inception breast models by @kaczmarj in #169
• add link to github repo in navbar by @kaczmarj in #174
• fixed togeojson and tosbu commands by @swaradgat19 in #176
• Gracefully handle missing backends by @kaczmarj in #177
• skip openslide/tiffslide tests if those backends not available by @kaczmarj in #180
• changed GeoJSON format by @swaradgat19 in #179
• clone batch_coords tensor to prevent Too many open files error by @kaczmarj in #182

Full Changelog: v0.4.0...v0.4.1

Version 0.4.0

Welcome to WSInfer version 0.4.0! This version has several important changes, including:

• Licensing WSInfer under Apache 2.0
• Using a new tissue segmentation algorithm (for this reason, outputs from this release may differ from outputs of previous releases, even though models remain unchanged)
• the use of torchscript models hosted on HuggingFace Model Hub. users can supply local torchscript models. we have discontinued the use of weights and timm architectures.
• support python 3.11
• deprecate support for python 3.7

See the changes below for more information.

Thanks very much to our contributors!

What's Changed

• add support for python 3.11 by @kaczmarj in #38
• [ADD] vgg16 tils model by @kaczmarj in #112
• add --segmentation-dir option with pre-existing tissue masks by @kaczmarj in #119
• [ENH] adds ability to run inference inside ROI by @kaczmarj in #114
• [ENH] use pytorch 2.0.0 as base image by @kaczmarj in #115
• replace prad label grade4+5 -> grade4or5 by @kaczmarj in #122
• fix bug in setting state dict from relative path by @kaczmarj in #123
• handle Path instances for Weights.file by @kaczmarj in #127
• fix auto-skip bug and gracefully handle failed patching by @kaczmarj in #130
• Support MPS on Apple Silicon by @petebankhead in #132
• REF: use torchscript models from huggingface hub by @kaczmarj in #144
• [REF] use custom patching implementation and use Apache 2.0 license by @kaczmarj in #153
• Remove model-specific dockerfiles + Remove ROI functionality by @kaczmarj in #155
• Fix: use pep 604 type annotations by @kaczmarj in #156
• Fix/setup.cfg by @swaradgat19 in #157
• added pyproject.toml file by @swaradgat19 in #158
• DOCS: update website, docstrings, and types for upcoming 0.4.0 release by @kaczmarj in #160
• [DOCS] add version dropdown and code block copy button by @kaczmarj in #161
• [ENH] rm pyyaml dep by @kaczmarj in #164
• ADD: calcuate mpp with openslide tiff tags by @kaczmarj in #166

New Contributors

• @petebankhead made their first contribution in #132
• @swaradgat19 made their first contribution in #157

Full Changelog: v0.3.6...v0.4.0

Version 0.3.6

Greetings wsinfer 0.3.6! This version contains several bugfixes and improvements to documentation. It also includes multi-gpu support, though this does not seem to produce speedups. Perhaps it will speed up inference if the filesystem is fast enough to serve up tiles.

As always, please submit a new GitHub issue is you need help, find a bug, or need a new feature.

What's Changed

• handle 'git' not being installed by @kaczmarj in #90
• [FIX] do not panic on unreadable files by @kaczmarj in #95
• [FIX] remove duplicate choices in 'wsinfer run' cli + clarify model-weight pairs by @kaczmarj in #96
• set cache_tilesource_maximum to 1 by @kaczmarj in #99
• write one run_metadata.json file per run by @kaczmarj in #100
• [ENH] add multi-gpu support by @kaczmarj in #103
• add more intro content to index.rst by @kaczmarj in #104

Full Changelog: v0.3.5...v0.3.6

Version 0.3.5

Many exciting changes in this release, including the new --speedup option to JIT-compile models prior to inference, better documentation, and more tests.

What's Changed

• ENH: add a --speedup option and tests by @kaczmarj in #75
• fix undefined path + add test of valid yaml config by @kaczmarj in #78
• place test_input on same device as model in jit_compile by @kaczmarj in #80
• allow pytest to fail in nightly by @kaczmarj in #84
• add cli docs with sphinx-click by @kaczmarj in #82
• divide by total time only if it is nonzero by @kaczmarj in #83
• print slide ids that failed patching or inference by @kaczmarj in #85
• use a default config for inception_v4 to prevent stderr message by @kaczmarj in #86
• [DOCS] fix readthedocs build by adding config by @kaczmarj in #87
• [CI] build docs + separate style/type tests by @kaczmarj in #88

Full Changelog: v0.3.4...v0.3.5

Version 0.3.4

The inceptionv4 models have been moved from a "custom" implementation to the timm implementation. This made it easier to convert the models to torchscript. This release also includes a script to convert the wsinfer models to torchscript. These models are included as assets in this release.

What's Changed

• Dockerfiles: use wsinfer version 0.3.3 by @kaczmarj in #63
• FIX: use inceptionv4 from timm by @kaczmarj in #69

Full Changelog: v0.3.3...v0.3.4

Version 0.3.3

This release fixes a bug in the installation of wsinfer. In version 0.3.2, installing wsinfer would not install a required data file. This has been fixed in 0.3.3. The 0.3.2 Docker images are not affected. Please do not pip install wsinfer version 0.3.2.

What's Changed

• ENH: add more sphinx documentation + update docker versions by @kaczmarj in #55
• REF: make modellib private + support numpy docstring style in sphinx docs by @kaczmarj in #56
• FIX: add geojson validation in tests + update copyright year by @kaczmarj in #59
• TST: run tests in docker image + fix setup.cfg regression by @kaczmarj in #61

Full Changelog: v0.3.2...v0.3.3

Version 0.3.2

We have a website now! This version updates the dockerfiles to use the new config specs. And many housekeeping things, like improved testing.

What's Changed

• ADD: tests of patching and of run_metadata.json by @kaczmarj in #49
• ENH: update dockerfiles and add version to config spec by @kaczmarj in #50
• ENH: update app dockerfiles + add mypy/flake8 to github ci + support weights from timm by @kaczmarj in #51
• FIX: use updated run_metadata.json keys in wsinfer tosbu by @kaczmarj in #52
• ADD: sphinx docs by @kaczmarj in #53
• DOCS: add usage examples and dedicated installation page by @kaczmarj in #54

Full Changelog: v0.3.1...v0.3.2