3D image segmentation

[AndreaPi]

Motivation Medical imaging is one of the main use cases for semantic segmentation (heck, U-Net was actually invented to perform Biomedical Image Segmentation!), and medical imaging today is all about 3D imaging. 3D image segmentation even has its own section on PapersWithCode, and there are competing libraries which allow 3D medical imaging segmentation. Thus, it would make a lot of sense to extend MMSegmentation to 3D semantic segmentation.

Related resources https://github.com/black0017/MedicalZooPytorch https://paperswithcode.com/task/3d-medical-imaging-segmentation https://arxiv.org/abs/1606.06650 https://arxiv.org/abs/1804.02967 https://arxiv.org/abs/1810.11654 https://arxiv.org/abs/1904.00625

[MeowZheng]

It is a very good proposal, and also in mmseg's backlog, but we suffer from a shortage of manpower, so 3d segmentation might not be supported in 2022. Any collaborations are appreciated If you are interested in it.

[FabianSchuetze]

What a wonderful idea! Are there any issue in the backlog related to that? What kind of support might be helpful?

[MengzhangLI]

What a wonderful idea! Are there any issue in the backlog related to that? What kind of support might be helpful?

You could list your requirement here and if you want to contribute to this repo you could directly open an issue or pull request and we would answer ASAP.

[AndreaPi]

Thanks @MengzhangLI , I added my requirements to the roadmap.

[FabianSchuetze]

@MengzhangLI : Thanks for your reply! How would you see the relationship with mmdetection3d which also contains a few segmentation models based on 3D data?

[thiagoribeirodamotta]

Are there any news on this matter?