License for the repo

[trathpai]

Thanks for the great work. Could you also tell which license will be applicable to the repo?

[fakerybakery]

+1!

[WongKinYiu]

I think it should be GPL3, I will check and update the license file.

[fakerybakery]

Any chance it could be permissively licensed?

[KleinYuan]

Would be great if you can make it MIT or Apache.

[nick-konovalchuk]

For now it's definitely at least GPLv3 😔
A lot of code smells ultralytics ish and classify directory even has YOLOv5 GPLv3 notices in files.
Is it possible to rewrite the code (even with some functionality loss), the that the repo is clean?

[Sharpz7]

I would be willing to help re-write the codebase to get it in a state it can be under MIT or Apache.

It would probably have to be a larger effort though - would anyone else want to help me? Would be great to get a state-of-the-art model under an relaxed open license.

[nick-konovalchuk]

I would love to help too, but we would definetly need some management and supervision from @WongKinYiu

[zeeshan3945]

Hi, I would love to contribute too. Is it possible if we can manage the environment to work on it.

[nick-konovalchuk]

Does anyone know if we can escape ulatralytics this way?

[WongKinYiu]

For me, the most hard part to re-implement is data loader and DDP training.
I am not familiar to write code for DDP training.
Architectures and training strategies are almost designed by ours team, there are no problems and issues to rebuild in new codebase.

[xenova]

Just to confirm, are the model weights licensed as GPL3 or AGPL3? As stated on https://www.ultralytics.com/license (FAQ) for YOLOv8:

Are Ultralytics YOLO-trained models licensed under the AGPL- 3.0 license?
Yes, all Ultralytics YOLO-trained models fall under the AGPL-3.0 License. The AGPL-3.0 License covers the training code and the models produced by that training code.

[WongKinYiu]

GPL3 license added.

[Sharpz7]

Yeah I'm pretty confident it has to be AGPL not GPL??

[nick-konovalchuk]

@Sharpz7 I'm quite sure this repo is mostly YOLOR, which was derived from yolov5 at the point when it was GPL. Thus, GPL

[KleinYuan]

For me, the most hard part to re-implement is data loader and DDP training. I am not familiar to write code for DDP training. Architectures and training strategies are almost designed by ours team, there are no problems and issues to rebuild in new codebase.

Employing Pytorch-Lightning, DDP is build-in. The code volume will also be largely reduced.

Exciting to see the re implementation efforts! @nick-konovalchuk

[Sharpz7]

@KleinYuan I mean this seems promising.

@WongKinYiu Can you please re-open this ticket? I want to pass this around more, because this is something the community could really do with.

[GrantorShadow]

I would be willing to help re-write the codebase to get it in a state it can be under MIT or Apache.

It would probably have to be a larger effort though - would anyone else want to help me? Would be great to get a state-of-the-art model under an relaxed open license.

Hi folks,
Could also be a part of this and contribute towards getting YOLOV9 to an MIT license? Would love to be a part and code towards this!

Thanks!

[VoNhatVinh]

I would be willing to help re-write the codebase to get it in a state it can be under MIT or Apache.

It would probably have to be a larger effort though - would anyone else want to help me? Would be great to get a state-of-the-art model under an relaxed open license.

Hi friend,
I would love to be a part this and under MIT lisense?

Thank you

[Sharpz7]

Hey All,

Please see our active efforts here:

• [EPIC] Scope out all work that needs to be completed YOLO#3

I am hoping to find time to work on this more extensively soon.

Thanks!