Changelog

Version 4.0.0 - 2022-02-03

View detailed changelog

NOTICE

• Due to the change in implementations on an upgraded dependency (skimage v0.14.2 -> v0.18.3), the morphometrics values “solidity” and “orientation” differ from values produced by AxonDeepSeg v3.x. “solidity” values produced can vary slightly due to numerical precision in algorithmic changes in skimage (on the order of 1%), a the convention for “orientation” changed from “rc” to “xy”, leading to a difference in value by pi/2. See this issue for more details: https://github.com/neuropoly/axondeepseg/issues/589

BUG

• Fix NaNs not appearing when generating morphometrics with the GUI View pull request

FEATURE

• Integrate IVADOMED into project View pull request

• Changed all instances of imageio_imread by ads_utils.imread View pull request

INSTALLATION

• Added Mac M1 compatibility View pull request

Version 3.3.0 - 2022-01-14

View detailed changelog

BUG

• Fix morphometrics error with ellipse axon shape View pull request

• Fix patches2im_overlap when width = height View pull request

• Patch myelin thickness calculation View pull request

• Bug fix due to h5py package version in fsleyes View pull request

• Pin h5py~=2.10.0 as a workaround for upstream Keras/TensorFlow issues View pull request

• Resolve bug that makes TEM data not segment well with multiple files View pull request

ENHANCEMENT

• Added the axon shape selection in the GUI (take two) View pull request

• Generate numbers image with CLI View pull request

• Refactor the code for the settings menu View pull request

• Add Optical Microscopy model in notebooks View pull request

• Add perimeter to morphometricsView pull request

• Add CLI support to generate morphometrics file. View pull request

• Replaced rcParams with font_manager to find the font View pull request

• Remove wildcard imports View pull request

• Migration to GitHub Actions View pull request

• Change the name of -o flag to -overlap flag View pull request

• Add a model to .gitignore View pull request

• Change the default name for saving the morphometrics file View pull request

• Use pathlib in the ads_plugin instead of os.path View pull request

• Integrate Simeon’s model with FSLeyes and CLI View pull request

• Fix Naming Convention View pull request

• Fixed Notebooks and Binder links in docView pull request

• Move test files to OSF, add download functionality View pull request

• Add postprocessing tests View pull request

FEATURE

• Implement Batch morphometrics for images present in directory/directories View pull request

• Implement Ellipse Minor Axis as Diameter View pull request

• Add a settings menu to the FSLeyes plugin View pull request

• Add Axon mask simulator View pull request

DOCUMENTATION

• Update Youtube video links View pull request

• Apply RTD fix to make builds pass again View pull request

• Update figure link View pull request

• Migrate images to new repo View pull request

• Change link View pull request

• Add description of morphometrics columns headings View pull request

• Add a reference to Readme.md View pull request

• Update the documentation for installing the GPU compatible version of ADS View pull request

• Add tutorial video link in RTD View pull request

• Fix docs for optical microscopy model View pull request

• Documentation updates View pull request

• Fix new changelog display in RTD View pull request

• Add zoomed window on masks examples in RTD View pull request

• Remove PyPI install instructions from RTD View pull request

• Fix youtube video link in RTD View pull request

INSTALLATION

• Link the test_files to GitHub View pull request

• Download models from Github repo View pull request

• Generate images synthetically in tests View pull request

• Change install to one env file & update docs View pull request

• Refactor AxonDeepSeg/models directory and update .gitignore View pull request

• Change how ADS dependencies are installed View pull request

• Add PR template View pull request

• Remove old config file (set_config) View pull request

Version 3.2.0 - 2020-10-16

View detailed changelog

BUG

• Fix redownloading models bug. View pull request

• [Bug] Resolve segment folder bug when using relative paths. View pull request

• Resolve bug that makes TEM data not segment well (#293 and #249). View pull request

• Add missing import to apply_model. View pull request

ENHANCEMENT

• FSLeyes plugin: default morphometrics file extension and font size . View pull request

• Remove gaussian blur option for data augmentation. View pull request

• Remove overlap when saving masks. View pull request

• Resume training from checkpoint. View pull request

FEATURE

• Add an “Axon numbers” overlay in FSLeyes. View pull request

• [FSLeyes plugin] changed Image.save/open with ads_utils.read/write. View pull request

DOCUMENTATION

• Add warning to doc for re-download of model folders when re-installing. View pull request

• Update documentation for manual masks creation. View pull request

• Add resampled_resolutions parameter in getting_started notebook. View pull request

• [Doc] Add notice about restarting FSLeyes to see the plugin for installation. View pull request

• Fix format in Changelog. View pull request

INSTALLATION

• Fix for the Windows installation. View pull request

• Move test images in default models folder. View pull request

• Upgrade to Python 3.7. View pull request

• Use a specific version of FSLeyes for the plugin. View pull request

Version [3.0] - 2020-03-13

Changed:

• Refractored data augmentation to use Albumentation library.

• Cleaned jupyter notebooks and reduced from 13 notebooks to 5.

• Switched to Dice Loss from Categorical Cross Entropy as loss function.

• Updated SEM and TEM models for better performance.

• Shifted AxonDeepSeg from TensorFlow to Keras framework.

• Upgraded CUDA to 10.0 and tensorflow to 1.13.1.

• Resolve image rescale warnings

• Handle exception for images smaller than minimum patch size after resizing

• Revert tensorflow requirekment to 1.3.0 and remove tifffile requirement

• Remove matplotlib.pyplot from source code and refactor to full OO plotting

• Standardize path management to pathlib library

• Shifted AxonDeepSeg from TensorFlow to Keras framework.

• Upgraded CUDA to 10.0 and tensorflow to 1.13.1.

• Add FSLeyes plugin

Version [2.1] - 2018-09-25

Changed:

• Fixed bug that would crash when user inputed consent for Sentry tracking

Version [2.0] - 2018-09-11

Changed:

• Upgraded ADS for Python 3.6-compatibility (no longer supporting Python 2.7)

• Minor changes to make ADS Windows-compatibile

• Removed plot hold commands (deprecated)

Version [1.1] - 2018-08-02

Changed:

• Minor Mac OSX-related bug fix

• Changed installation requirements to exact release versions

Version [1.0] - 2018-08-02

Versions 1.x will remain Python 2.7-compatible

Version [0.6] - 2018-08-01

(version 0.5 was skipped due to conflicting file on PyPI)

Added:

• Comprehensive testing suite

• Bug tracking (Sentry)

• Blue-red visualisation function for segmented masks

Changed:

• Dataset building and training notebook

• Minor documentation improvements

• Minor bug fixes

Version [0.4.1] - 2018-05-16

Added:

• GIMP procedure for ground truth labelling or segmentation correction added in the documentation.

• Compatibility with tiff images.

• Continuous integration with Travis is now supported.

Changed:

• The documentation website is now hosted on ReadTheDocs.

• Updated documentation on the usage of AxonDeepSeg.

• Change of axon and myelin masks filenames for better clarity.

Version [0.3] - 2018-02-22

Added:

• Compatibility for image inputs other than png

• Pre-processing of input images is now done inside AxonDeepSeg

Changed:

• Help display when running AxonDeepSeg from terminal