This archive stores hyperparameter files, parameter files, test-set predictions etc. for team Multi-Planar U-Net's submission to the 2019 "The International Workshop on Osteoarthritis Imaging Knee MRI Segmentation Challenge".
The challenge and results are described in: https://arxiv.org/abs/2004.14003

Our team fitted 7 instances of our model to 7 random splits of the training data.
Each of those 7 models may be used to predict knee MRIs in .nii.gz format using the Multi-Planar U-Net software package.

* Please refer to the guide at https://github.com/perslev/MultiPlanarUNet for further details on the mpunet software package.
* Please refer to the challenge paper for details on the dataset that the 7 models were trained on.

The following section shows the basic usage of our software and pre-trained models to predict on a new knee MRI file:

'''
# Install the mpunet software package
git clone https://github.com/perslev/MultiPlanarUNet
pip install MultiPlanarUNet

# Predict on file 'my_mri.nii.gz' using the model of the first dataset split ('split_0')
mp predict -f ./path/to/my_mri.nii.gz --project_dir models/split_0 --out_dir my_predictions --no_eval
'''