I could not really find a satisfying answer here yet, this question touches it but kind of reversed.

  1. For example the training framework for Yolo v7 is used to train a neural network using the config (cfg) file from the repo which is licensed under GPL-3 using custom data.

  2. Now the model gets exported to onnx (or tflite or any other major inference runtime) and used with inference scripts that are completely self-written (inference done by onnxruntime, especially post-processing like decoding and nms self-written).

Is (2) now derivative work and must be distributed under GPL-3 too by using the neural network weights?

  • 8
    The non-answer to this question is that there is no clear answer to this yet. It's going to be an evolving area of copyright law over the next few years. Commented Aug 17, 2022 at 19:55
  • 4
    It is not really answerable legally. I suspect it is probably going to be regarded as output data of a GPL program, just like mentioned in the question opensource.stackexchange.com/questions/5478/… The other possibility, as mentioned in the first question you linked to, is that the model data are not even copyrightable, as they were not created by a human.
    – Brandin
    Commented Aug 18, 2022 at 8:03
  • 2
    Thanks to both of you, that is what I thought after not finding a lot of resources on this matter. Interestingly, some Kaggle competitions do not allow YOLO models because of the GPL 3 License. Also the creator of Yolo v3 and v5 does not really have an answer. Maybe, I compile some results of my search on this later when I have time. Many thanks so far!
    – sofa28
    Commented Aug 18, 2022 at 12:30


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.