Segmentation of CT scans with use of Pytorch Lightning and segmentation models pytorch.
Install this directory with pip install -e .
.
Create folder input
in this directory.
Download https://www.kaggle.com/datasets/andrewmvd/covid19-ct-scans or similar dataset and unzip it into input
as covid19-ct-scans
(so input/covid19-ct-scans/metadata.csv
is relative path to medata).
Run python ./scripts/prepare_pngs.py
to preprocess dataset into 256*256 images.
Run python ./scripts/train.py
to start training with basic configuration. There are placeholder, that reduces size of dataset to 100 pngs and training goes for 1 epoch.
Result of training can be viewed in ./notebooks/visualization.ipynb
(requires jupyter notebook, pip install notebook ipywidgets
)
Uninstall this directory by pip uninstall ct_covid_segmentation
as you done with it