-
Notifications
You must be signed in to change notification settings - Fork 22
Query regarding model licensing #444
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
Thank you @ThibautGoldsborough for raising this issue. |
short update: we did not find time to discuss this this week and take a holiday break until January... |
short update: raised in today's meeting. We are now working on defining the scope of the |
we have (finally) discussed the important issue you raised. We would like to propose the following:
The next version of our model format should additionally be changed such that listed datasets also make the associated license explicit. This is not currently possible and a workaround would be to use any of the optional comment fields. Please let us know if you have diverging opinions or if you want to discuss any other aspects regarding this topic. Thanks a lot, |
For datasets that require derivative works to be shared under a particular type of license, this would seem to preclude sharing using the bioimageio format? It's also unclear, if the license applies only to the metadata, under which license all other files distributed by bioimageio fall under? And therefore the use of any parts of a bioimageio model becomes legally unclear |
Hello,
Please feel free to redirect me to the Image.sc forum if this is not the right place to discuss about licensing.
We are in the process of uploading a number of instanseg models and had a query regarding the licensing of our models. The instanseg method is released under Apache-2.0, however the weights were trained on a number of datasets, each with their own licensing terms. So far we have kept to CC-BY or CC-0 datasets as these are compatible under the Apache-2.0 terms. But we'd like to release more models trained on less permissive licences (e.g. cellpose dataset, tissuenet dataset, livecell dataset), which have confusing strictly non-commercial custom licenses (cellpose and tissuenet).
I noticed that some of the popular models hosted on bioimageio (e.g. https://bioimage.io/#/?id=10.5281%2Fzenodo.5869899&type=model) were released under a fully open license (CC-BY-4.0) yet the training data (livecell) is released under a more restrictive CC-BY-NC. Does bioimageio assume that the license of a model is not tied to the license of the training dataset?
I'm aware this is not your direct responsibility, but there seems to be a lack of guidelines for model developers which could lead to unintended downstream risks for model users.
The text was updated successfully, but these errors were encountered: