Datasets:

Languages:
English
ArXiv:
License:

🚩 Report : Legal issue(s)

#19
by RDDjensen - opened

The Objaverse dataset was scraped from Sketchfab without their knowledge, and without consent of many of the artists whose work it features.
It doesn't even take into account Sketchfab's "NoAI" tag, as many artists have found their models in this dataset without their consent.
Matt Deitke and his team are opening themselves up to a lawsuit if they don't rebuild this dataset from scratch with assets that are exclusively from artists who have EXPLICITLY OPTED INTO AI TRAINING.

SketchFab's statement: https://twitter.com/Sketchfab/status/1639311132253995023
Affected artists: https://twitter.com/AustinBeaulier/status/1639073980618522625
https://twitter.com/miguelbandera/status/1639015194729824256
https://twitter.com/That_GuyMike/status/1639189771992936450

mmw: they're going to throw out RDDjensen's case, and RDDjensen and co. will be absolutely devastated by a defamation that sticks.

Jezus Christ, how many models DID you guys rip from Sketchfab huh?

160 pages x ~ 3000 models = ~480k !?!
Really?!

According to the CEO, this scrape was done before the no-AI system was implemented. https://twitter.com/albn/status/1639182560621211650?s=20

no ai tags are meaningless. Sketchfab is a for profit company with databases full of ripped assets. Bunch of cry hards calling something theft without even knowing what the definition of theft is. Maybe stop publishing your stuff publicly on sites that convert your work to creative commons then making up incoherent legal theories. Luddites are so cringe.

It baffles me how they think it's fine to use your assets for machine learning without your permission or even letting you know, with the ultimate goal of replacing you. You have to do some mental gymnastics to think that.

@Entity I'm not a lawyer, but I think you can consider the No AI tag to be an extension of the license, and as such, it should be quite meaningful.

I'm a bit confused though by your statement about Sketchfab being a for profit company with databases full of ripped assets. Do you think that makes it more right to download and use their models? Using digital content that you know has been stolen is most likely in itself illegal, no matter whether there is any No AI tag on it or not.

@Eddy22 I'm not sure you understand how machine learning works. You always create as large datasets as possible. The dataset description says "Objaverse is a Massive Dataset with 800K+ Annotated 3D Objects."

Sign up or log in to comment