Submitting a model without a dataset

[snakers4]

Hi!

We would like to contribute our models.

PR Rejection
Your PR may be rejected in some cases, for example:
If a dataset is inaccessible.

For a number of legal reasons we are not planning to share our datasets and / or re-upload publicly available datasets (i.e. LibriSpeech derivatives).

Does this make contributing a model to your model zoo improssible?

[vladimir-dudnik]

@snakers4 please review CONTRIBUTING.md for terms we require to accept contribution.
Contribution not required to re-upload publicly available dataset, we ask to specify which dataset your contribution depends on and what is the dataset license.

[snakers4]

@vladimir-dudnik

we ask to specify which dataset your contribution depends on and what is the dataset license.

And if the dataset in question is private?
Does it imply that the contribution will be automatically rejected?

[vladimir-dudnik]

@yssemaev could you please review question from legal point of view?

[yssemaev]

Hi, we accept contribution according to CONTRIBUTING.md "under permissive licenses, such as MIT, Apache 2.0, and BSD-3-Clause" only. As a rule, in github repo it's mentioned which dataset was used for training and validation (public or internal). Only contribution of a model is accepted, not a dataset. OMZ's Apache 2.0 license allows various use of models including commercial or academic.

[snakers4]

under permissive licenses, such as MIT, Apache 2.0, and BSD-3-Clause

Is GPL considered to be a permissive license?

[yssemaev]

No, L/GPL implies additional restrictions like "same license", etc.

[snakers4]

I see.
Many thanks.