detection ensemble prediction

[vankhoa21991]

Hello, in the detection example, there is no ensembling of the predictor to make prediction. This feature is going to be added in the future?

[Can-Zhao]

Thank you for the feedback! Would you be interested in contributing a PR for it?

[vankhoa21991]

I already made some adjustment for that, using the library https://github.com/ZFTurbo/Weighted-Boxes-Fusion to ensemble box predictions from different folds. However I haven't made a PR before, if you give me instruction I can make one, thanks for your help.

[Can-Zhao]

That's awesome! The complete document of making PR is here: https://github.com/Project-MONAI/MONAI/blob/dev/CONTRIBUTING.md

This is my own cheatsheet to upload PR, you would need to name a $PR_BRANCH for your PR.
$local_monai_dir is the path of your local git code of MONAI.

create branch:
(1) git checkout -b $PR_BRANCH
(2) make edits at $local_monai_dir

reformat your scripts:
(3) export PYTHONPATH=${local_monai_dir}
(4) run cmd: runtests.sh --autofix
or run isort, black, flake8 one by one
(5) mypy ${your_added_script}
make edits to pass all the tests above

(6) submit PR, run cmd:

find . | grep -E "(__pycache__|\.pyc|\.pyo$)" | xargs rm -rf
git add .
git commit -a -s -m "what changes you have made"
git push origin $PR_BRANCH