Sharing your models on the Hugging Face Hub about goal

[otaheri]

Hi @NimaBoscarino , thank you for reaching out regarding publishing our work on the great hugging face space.

Definitely, I would be interested in adding this work and our two other works on hugging face. However, there are some challenges in this regard like 1. The license for our body/hand models (SMPLX and MANO models), 2. The license for the 3D objects in our GRAB dataset, and 3. The trained models. I am sure you have a solution for these :). If so, I would be more than happy to start adding them on hugging face and would appreciate your help along the way.

Please let me know if you have any solutions and then I will get my hands on moving to hugging face hub ;).

Best,
Omid

[NimaBoscarino]

Hi @otaheri, apologies for the delayed reply! Glad to hear you're interested in publishing your work on Hugging Face 😄 As for the challenges:

1. For models I don't think you would need to worry about the licenses of the dependencies since you'd only be uploading your model's weights, but if you wanted to make a demo on Hugging Face Spaces it could be something to consider. For SMPLX, for the Yuliang/ICON space it looks like the author added it directly into their code repo. In your code, you install it in the requirements.txt which is also good! There would be no need to upload SMPLX to the Hugging Face Hub, and users on the Hub won't be able to directly access that model's checkpoints, so (from the way I understand SMPLX's license) I think that it would be okay as it is. Likewise for the MANO models!
2. We have a number of licenses included in the Hub, and you could also specify your own license like you've done on GitHub! If you'd like to add further restrictions on how the GRAB dataset can be accessed, you could do it by making it a "gated" dataset, or you could even make it a private dataset if you only want to share it with other members of your research group.
3. Same as with datasets, you can add your own license, and even restrict the model as a "gated" model or a private model.

Hopefully that helps a bit! Let me know if there are still open questions 😄

Also, if you'd like, I can create a "Max Planck Institute for Intelligent Systems" organization on Hugging Face so that all your material can live in there!

[NimaBoscarino]

Hey @otaheri! Just bumping this in case you're still interested 😄 I'm happy to help out with it!