Support for OpenLabel standard

cvat-ai / cvat

Annotate better with CVAT, the industry-leading data engine for machine learning. Used and trusted by teams at any scale, for data of any scale.

https://cvat.ai

MIT License

12.34k stars 2.97k forks source link

Support for OpenLabel standard #3999

Open dnns92 opened 2 years ago

dnns92 commented 2 years ago

My actions before raising this issue

[ X] Read/searched the docs
[ X] Searched past issues

Dear CVAT-Team,

there have been some efforts to unify labeling standards into one big standard: It's called openLabel. And I think CVAT should support it.

You can find the concept paper here: https://www.asam.net/index.php?eID=dumpFile&t=f&f=3876&token=413e8c85031ae64cc35cf42d0768627514868b2f

I think it would make sense to integrate most parts of it to cvat, e.g. object detection, classification, scenarios, etc.

Why should you support it?

It's open
It's generic
It's well described
It will be (I think) an accepted standard to label objects, classify Images, etc.

Cheers

nmanovic commented 2 years ago

@dnns92 , thanks for the feature request. First of all it is a feature request for Datumaro project: https://github.com/openvinotoolkit/datumaro. All formats in CVAT are supported using the dataset framework. It is developed by our team as well.

Which known datasets support the format? Do you know about any SDK which can read the format already? Please let me know if you can contribute and send us a PR with the format. We we are ready to discuss, review, provide a feedback and merge it.

dnns92 commented 2 years ago

Thank you for the reply. So far, I do not think a commonly used dataset/SDK that supports it. I guess its a hen-egg problem. Its a very generic standard, that can do more than e.g. coco does. Unfortunately I do not have the spare time to provide a pull request.

Cheers

timheb commented 6 months ago

Maybe this library can be used. Unfortunately no update for 3 years. https://github.com/Vicomtech/video-content-description-VCD