Depth anything v2 small weights has Apache 2.0
Can you please let me understand, what is the data set that it is trained on?
If it is trained on cc by nc 4.0 dataset, then the model is also cc by nc 4.0.
Or
Is Depth anything v2 small trained on only cc by 4.0?
Originally posted by @codimahi in #81
Background
Regarding the discussion in issue #81, I noticed that Depth-Anything-V2-Small is released under the Apache-2.0 license, while other larger versions (Base, Large, Giant) are under CC-BY-NC-4.0.
In the previous discussion, it was mentioned that the model size affects the license due to the dependencies or datasets involved. I would like to clarify the commercial safety of the Small model more specifically.
Questions
Does the Depth-Anything-V2-Small model only use datasets that are fully cleared for commercial use?
Even if we don't know the full list of datasets, has the team ensured that any "license-commerciality" issues related to the training data have been resolved for the Small model so that end-users do not inherit any license risks when using it commercially?
Specifically, if a user uses the Small model weights for a commercial service, can they be certain that no part of the training process (e.g., specific labeled datasets with restrictive licenses) limits this commercial application?
I am asking this because while the model weight itself is Apache 2.0, the "data-derived" license risk is a critical concern for commercial deployment.
Thank you for your great work and for providing this lightweight model to the community.
Originally posted by @codimahi in #81
Background
Regarding the discussion in issue #81, I noticed that Depth-Anything-V2-Small is released under the Apache-2.0 license, while other larger versions (Base, Large, Giant) are under CC-BY-NC-4.0.
In the previous discussion, it was mentioned that the model size affects the license due to the dependencies or datasets involved. I would like to clarify the commercial safety of the Small model more specifically.
Questions
Does the Depth-Anything-V2-Small model only use datasets that are fully cleared for commercial use?
Even if we don't know the full list of datasets, has the team ensured that any "license-commerciality" issues related to the training data have been resolved for the Small model so that end-users do not inherit any license risks when using it commercially?
Specifically, if a user uses the Small model weights for a commercial service, can they be certain that no part of the training process (e.g., specific labeled datasets with restrictive licenses) limits this commercial application?
I am asking this because while the model weight itself is Apache 2.0, the "data-derived" license risk is a critical concern for commercial deployment.
Thank you for your great work and for providing this lightweight model to the community.