liver-ct-segmentation-package 1.6.0

Prediction package for U-Net models trained on the LiTS dataset.

liver-ct-segmentation-package

Prediction package for reproducible U-Net models trained on the LiTS dataset. These models are trained using the mlf-core framework (https://github.com/mlf-core/liver-ct-segmentation/).

• Free software: MIT

• Documentation: https://liverctsegmentationpackage.readthedocs.io.

Features

• Model download CLI: liver-ct-seg-model-dl

• Prediction CLI: liver-ct-seg-pred

• Uncertainty of prediction CLI: liver-ct-seg-uncert

• Input feature importance (Guided Grad-CAM) CLI: liver-ct-seg-feat-ggcam

Credits

This package was created with mlf-core using cookiecutter.

Changelog

This project adheres to Semantic Versioning.

1.6.0 (2021-11-01)

Added

• Introduced command to compute uncertainty of prediction, via the monte-carlo dropout procedure.

• Introduced command to compute input feature importance via the guided grad-cam algorithm, as implemented by the captum lib.

Fixed

Dependencies

Deprecated

1.5.0 (2021-09-01)

Added

• First working prototype

Fixed

Dependencies

Deprecated

0.1.0-SNAPSHOT (2021-07-28)

Added

• Created the project using mlf-core

Fixed

Dependencies

Deprecated