Code release for the paper Zero-Shot Domain Adaptation in CT Segmentation by Filtered Back Projection Augmentation
git clone https://github.com/STNLd2/FBPAug.git
./FBPAug/install.sh
from fbp_aug import apply_conv_filter
sharper_ct_image = apply_conv_filter(ct_image, a=30, b=3)
To run a single experiment please follow the steps below:
First, the experiment structure must be created:
python -m dpipe build_experiment --config_path "$1" --experiment_path "$2"
where the first argument is a path to the .config
file e.g.,
"~/FBPAug/configs/expepiments/baseline.config"
and the second argument is a path to the folder where the experiment
structure will be organized, e.g.
"~/miccai_paper/baseline"
.
Then, to run an experiment please go to the experiment folder inside the created structure:
cd ~/miccai_paper/baseline
and call the following command to start the experiment:
python -m dpipe run_experiment --config_path "../resources.config"
where resources.config
is the general .config
file of the experiment.