Pretrained models

Hello Organizers,

Can we use pretrained models trained with medical image datasets (non-imagenet datasets) ?

Regards,

Yes, this is allowed. Do note that if you use such models, they should be available under a permissive licence. This is because the code for containers submitted to the final round needs to be released under a permissive open-source license. Also note that when your code is running on Grand Challenge, it does not have internet access, so it cannot download pretrained models. Any pretrained models should therefore be stored in your submitted container.

I hope this answers your question, Luuk

Hello,

What do you mean by permissive licence ? is there a specific list of licences we are allowed to use (e.g. MIT, Apache 2.0...) or as long as anyone can download the model / software it's okay (e.g. a model released on a GitHub repo)

Edit : from Wikipedia https://en.wikipedia.org/wiki/Permissive_software_license, it means MIT, Apache, BSD and CC-BY

Thank you, Simon

Hi,

This is the list of open source licences that Grand Challenge automatically recognizes when submitting a Github repository as an Algorithm: Apache License 2.0 MIT License GNU GPLv3 GNU AGPLv3 GNU LGPLv3 Mozilla Public License 2.0 Boost Software License 1.0 The Unlicense

If you prefer another permissive licence, just let us know. We'd be happy to discuss.

Kind regards, Luuk