BONBID-UNet-3D Segmentation Algorithm Docker

UNet-3D network used in BONBID-HIE challenge.

Prepared by Gani Rahmon, 9/20/2023

Before cloning

This project uses Git Large File Storage (LFS) to handle large files. To work with this repository properly, you'll need to have Git LFS installed and set up. Here's how:

1. Install Git LFS

First, ensure you have Git LFS installed. If not, download and install it from https://git-lfs.github.com/.

2. Clone the Repository

If you haven't cloned the repository yet, you can do so with:

git clone <repository-url>

If you've already cloned the repository before installing Git LFS, navigate to your repository and pull the LFS files:

git lfs install
git lfs pull

Install requirements

pip install -r requirements.txt

Run with Docker

Before running, you will need a local docker installation. For more details, please read grand-challenge documents https://grand-challenge.org/documentation/automated-evaluation/ and https://comic.github.io/evalutils/usage.html

predict.py is the main function for generating prediction results.

Noted that when your algorithm is run on grand challenge <local_path>/case1 will be mapped to /input. Then a separate run will be made with <local_path>/case2 mapped to /input. This allows grand challenge to execute all the jobs in parallel on their cloud infrastructure. For simplicity, you can include one case in your test data when you test locally. The platform will handle multiple cases. Predict should only handle one case.

Please follow these steps to run it on the local machine.

1. Build the docker

./build.sh

2. Test the docker

./test.sh

In test.sh, use the following command in order to generate the results locally and test your codes:

docker run --rm \
    --memory="${MEM_LIMIT}" \
    --memory-swap="${MEM_LIMIT}" \
    --network="none" \
    --cap-drop="ALL" \
    --security-opt="no-new-privileges" \
    --shm-size="128m" \
    --pids-limit="256" \
    -v $SCRIPTPATH/test/:/input/ \
    -v $SCRIPTPATH/output/:/output/ \
    bondbidhie2023_algorithm

But for uploading algorithm docker to the grand challenge server, please use the codes that I provided in test.sh.

docker run --rm \
    --memory="${MEM_LIMIT}" \
    --memory-swap="${MEM_LIMIT}" \
    --network="none" \
    --cap-drop="ALL" \
    --security-opt="no-new-privileges" \
    --shm-size="128m" \
    --pids-limit="256" \
    -v $SCRIPTPATH/test/:/input/ \
    -v bondbidhie2023_algorithm-output-$VOLUME_SUFFIX:/output/ \
    bondbidhie2023_algorithm

3. Exporting docker ./export.sh Running ./export.sh, and submitting the generated zip file of the algorithm docker.

References

1. Original repository by Rina Bao: https://github.com/baorina/BONBID-HIE-MICCAI2023
2. Challanege website
3. Dataset download
4. Preprint challenge paper
5. MONAI
6. Hausdorff Loss