Hello, I’m considering using CC5 as a source of training data for some machine learning models I’m working on. My pipeline would be something like this:
export character from CC5 to blender
render character with blender
use rendered images as training data
I’m having a hard time determining if this as an allowed usage. In the EULA [0] it appears that only ‘Content’ and ‘Derived Content’ is prohibited from being used in machine learning, and that images rendered from blender do not fit in either of these categories. This seems to be corroborated by a forum post from a couple years ago [1].
To be clear, I’m not attempting to replicate any CC5 functionality: the types of applications I wish to target are based around image segmentation.
Can anyone offer any clarification or advice about this?
Generally the EULA’s allow commercial use in video and image form, however there are a few restriction that are still stipulated. One important restriction is that the images or video can’t be part of a stock library, and another is they can’t be used for machine learning.
So sadly what you suggest would appear to be classed as derivative content which is part of the restrictions below.
(C) You may NOT use the Content or any Derivative Content in the following scenarios or applications:
As purchasable in-software content or downloadable paid content.
For online or in-software character generation.
In character generation APIs or API licensing.
As embedded content within applications or online services.
For machine learning, AI training, or AI-generated output.
3D Print for sale or for massive redistribution; create a mold for mass production.
You may not use the content in any pornographic and illegal use, or otherwise unlawful manner.
However, if you do feel that your proposed workflow doesn’t breach our EULA, then please feel free to contact Customer Support who will seek clarification for you and give you a definitive answer.
My question is really whether a rendered image is ‘Content’ or ‘Derivative Content’. Here are how the two terms are defined in the EULA:
1.1 Content - Refers to digital assets published at Reallusion’s Stores, including but not limited to 3D models, textures, materials, motions, audio, AML script, BVH, FBX, OBJ, images, image sprites, and any files in Reallusion proprietary formats.
1.3 Derivative Content - Refers to any content that is modified from Base Content and republished at Reallusion’s Stores by Vendor. Derivative Content must have significant modifications to the Base Content in order to be published for sale in Reallusion’s Stores.
I don’t quite see how rendered images fit either of these two definitions. But if you are not in a position to provide an official answer I probably should just reach out to customer support directly.