Looking forward exporting Mask2Former to ONNX

[Kenneth-X]

Describe the feature

Motivation

1. Many detection & sematic segmentation classical methods（e.g yolo,centernet, deeplabv3,ocrnet）has already support exporting to onnx and made a huge contribution for academia and industry field. But for panoptic segmentaion, mmlab series are still not ready for that yet
2. sota mask2former is a great achievement for panoptic segmentaion( e.g transformer) . It will be wonderful if can be export to onnx（swin transfomer backbone） for depolyment

Related resources

If there is an official code release or third-party implementations, please also provide the information here, which would be very helpful.
https://github.com/facebookresearch/Mask2Former
https://github.com/open-mmlab/mmdetection/tree/master/configs/mask2former

[M-youyou]

Thanks for your advice. Looking forward exporting Mask2Former to ONNX, also.

[lvhan028]

Totally agree with the motivation. Transformer models will be our top priority in April.

[lvhan028]

This feature has been put into our backlog, which is supposed to be released in v0.6.0

[p890040]

This feature has been put into our backlog, which is supposed to be released in v0.6.0

Hi, @lvhan028
Thanks to the efforts of the mmopen team
The v0.6.0 version has been released, but I noticed that there is still no Mask2former feature.
Is there any difficulty in converting Onnx to Mask2former again? Or do you need additional operation support?

Looking forward to your team's reply Thanks!

[lfxx]

This feature has been put into our backlog, which is supposed to be released in v0.6.0

Any update on this?hope for onnx on Mask2Former.

[eliabruni]

I am also very interested, any updates? Thank you!

[fernandorovai]

Hello, any news on converting the mask2former? It opens the possibility to deploy the model in large scale. Is there any way we can help?

[nubertj]

same here, we would also really appreciate this feature
@Idate96

[xunleiw]

Hello, is there an ETA when Mask2Former to ONNX feature will be supported?

[Robotatron]

any update?

[CristoJV]

any update now?

[nubertj]

@lvhan028 could you please reopen this issue?
Thank you :)

[lvhan028]

@RunningLeon may help handle this feature

[RunningLeon]

@Kenneth-X @nubertj @CristoJV @Robotatron @xunleiw hi, guys. If you have time, pls. try this PR #2347. Any comment on the PR is greatly appreciated.

[nubertj]

@RunningLeon amazing, thanks a lot.
We will give it a try :) M2F is still heavily used by us.

[github-actions]

This issue is marked as stale because it has been marked as invalid or awaiting response for 7 days without any further response. It will be closed in 5 days if the stale label is not removed or if there is no further response.

[github-actions]

This issue is closed because it has been stale for 5 days. Please open a new issue if you have similar issues or you have any new updates now.

[nubertj]

@pascal-roth could successfully deploy it on our end, thanks a lot for the great contribution, @RunningLeon and @lvhan028! :)