Open yewnyx opened 2 years ago
My background is a member of https://github.com/omigroup/omigroup, https://github.com/godotengine/godot and an advisor to Khronos.
An open-source and remixable motion dataset licensed similar to CC-BY-4 is essential to open research and development.
The established developers can afford their million-dollar labs and actors to create motion datasets. See Meta, EA, Ubisoft, and others. However, the students, practitioners, and hobbyists can't produce motion datasets casually.
See also: https://github.com/Miraikomachi and https://vrm-consortium.org/en.
We also wrote a VRM plugin that is MIT licensed opensource. https://godotengine.org/asset-library/asset/964 and would like to make future character motion research and development easier.
The dataset is in BVH format. Is there any chance to make the raw mocap marker data available? People have been asking for this in SMPL-X format. With the marker data, we can convert it to SMPL/SMPL-X using MoSh. Would you have any interest in including this in the AMASS dataset? You could keep the same NC license since most of AMASS's sub-datasets are also non-commercial.
Hi, Can I use Bandai-Namco-Research-Motiondataset to drive my mixamo skeleton human model?
The main thrust of this issue is that NC is not an open source license, and unsafe for use.
I would appreciate if requests unrelated to the top-line issue were politely filed in separate issues, please?
My concern is that even CC's general counsel and policy director could not give a clear statement on whether using such content for machine learning is legal or not:
https://creativecommons.org/2021/03/04/should-cc-licensed-content-be-used-to-train-ai-it-depends/
There is no consensus on whether the use of copyright works as input to train an AI system is an exercise of an exclusive right.
There remains significant legal uncertainty about whether copyright applies to AI training, which means it may not always be clear whether a CC license applies. In other words, there is no consensus on whether the use of copyright works as inputs to train an AI system is an exercise of an exclusive right (e.g., reproduction, adaptation, etc.).
Unless the main purpose of this public dataset is to allow third-party to verify Bandai Namco's own research (as in, reproducibility), I would say an alternative license that clear up such doubt would be more advisable.
(Ubisoft's dataset is licensed under the same term, but so far I am not aware of non-Ubisoft researchers using its dataset directly.)
The icense was changed to allow modification.
We do not plan to change it to commercial use, but we are preparing a separate license for commercial use.
With your new license, would you be interested in having us create a version of the dataset in SMPL-X format using MoSh++? If we can add this to AMASS, this would increase the use of the dataset.
hi @m3-takahashi - is there an update on the separate commercial license use?
I haven't looked at any of the data yet, because I saw the license and looked it up - CC-BY-NC-ND is "look but don't touch" closed-source. Concretely, it is: a) not open source, and b) considered to be dangerous ("NC option is a dangerous trap") [1] by the Open Source Initiative [2], who define the rules for what validly constitutes open-source (the "Open Source Definition" [3]).
I plead with you to please reconsider your choice of license.
Thank you, Yewnyx
[1] "CC-NC considered harmful" http://esr.ibiblio.org/?p=4559 - by https://en.wikipedia.org/wiki/Eric_S._Raymond [2] https://en.wikipedia.org/wiki/Open_Source_Initiative [3] https://en.wikipedia.org/wiki/The_Open_Source_Definition