Roadmap of MMSegmentation

[hellock]

We keep this issue open to collect feature requests from users and hear your voice. Our monthly release plan is also available here.

You can either:

1. Suggest a new feature by leaving a comment.
2. Vote for a feature request with 👍 or be against with 👎. (Remember that developers are busy and cannot respond to all feature requests, so vote for your most favorable one!)
3. Tell us that you would like to help implement one of the features in the list or review the PRs. (This is the greatest things to hear about!)

V0.6 (August)

• [x] ResNeSt. (#47)
• [x] Semantic FPN. (#35)
• [x] PointRend. (#35)
• [x] FastSCNN. (#58)
• [x] DNL. (#37)
• [x] ONNX export. (#12)

[xvjiarui]

• [x] fp16 in #18

[xvjiarui]

albumentation in #29

[zjupx]

MobileNet v3 backbone support please！

[VincentZ-2020]

Add U-net please！

[zhongqiu1245]

nnU-net support please！

[xungeer29]

tensorboard support please!

[xvjiarui]

tensorboard support please!

You may uncomment this line to enable tensorboard.

[JoyLuo]

Gans for segment

[advaza]

Is there a way to view segmentation results throughout training (like evaluation hook but with visualisation of possibly unlabelled dataset)? E.g. Tensorboard image writer.

[sandylaker]

UNet, UNet++, and installation on OS X.

[johnzja]

MobileNet v3 backbone support please！

Supporting...

[hhaAndroid]

support EfficientFCN

[JoyLuo]

dice loss， focal loss

[caodroid]

please add Res2net，thank you，

[jianlong-yuan]

please add Panoptic-DeepLab and segfix, thank you

[huang45]

https://github.com/PingoLH/FCHarDNet please!

[huang45]

and Bisenetv2, thank you~

[Hr-Song]

Lovasz hinge loss and Lovasz softmax loss suport please， thank you.

[ywwwer]

Multi-scale attention https://github.com/NVIDIA/semantic-segmentation, thank you!

[zhongqiu1245]

HighRes2DNet(2D version of HighRes3DNet) support please!

[zhongqiu1245]

support CaNet please，thank you！ https://github.com/icoz69/CaNet?utm_source=wechat_session&utm_medium=social&utm_oi=980147122032807936

[caodroid]

hope the Xception will be supported in backbones, thank you https://arxiv.org/pdf/1610.02357.pdf

[9p15p]

support COCO Dataset, please. Thank you.

[JPLAY0]

Support Mapillary Vistas Dataset ,thank you.

[zhongqiu1245]

Support Dice Loss, Generalized Dice Loss, Focal Loss & Tversky Loss please! Thank you!

[NickChang97]

Make it easy to use mixed loss

[lucci17]

please support for temporal structure

[caodroid]

DenseASPP plz，thank you

[JPLAY0]

Support CamVid Dataset ,thank you.

[jfzhuang]

Xception backbone support please！

[fcakyon]

@hellock please support VNet and ResUNet

[aikeyz]

EMA/SWA and knowledge distillation please

[zhongqiu1245]

Hi, thanks for your amazing job！ Is there any plan to support Panoptic-DeepLab？ Thank you！

[hannes09]

Hi, great project. I would like to see DDRNet. https://github.com/chenjun2hao/DDRNet.pytorch

[tiwaria9034]

Thanks for your contributions. Please add EfficientNet family of backbones (v1, v2, tflite, etc.)

[Junjun2016]

Thanks for your contributions. Please add EfficientNet family of backbones (v1, v2, tflite, etc.)

Could you please provide some references?

[tiwaria9034]

Thanks for your contributions. Please add EfficientNet family of backbones (v1, v2, tflite, etc.)

Could you please provide some references?

EfficientNet (v1):

• https://arxiv.org/abs/1905.11946
• https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/efficientnet.py

EfficientNet (v2):

• https://arxiv.org/pdf/2104.00298.pdf
• https://github.com/google/automl/tree/master/efficientnetv2

[eugenelawrence]

MaskFormer

[eliabruni]

Hi, great project. I would like to see Mask2Former

[litaws]

Add support for Un/Semi-Supervised Loss (-or- partially unlabeled data)

edit: moved into a separate issue #1147

[wmkai]

Hi, thanks for your great job！ Is there any plan to support Mask2Former？ Thank you！

[jyang68sh]

Please update PIDnet as it is the new sota

[AndreaPi]

Add 3D image segmentation capabilities #2010 In particular, I would propose to implement the following models:

https://arxiv.org/abs/1606.06650 https://arxiv.org/abs/1804.02967 https://arxiv.org/abs/1810.11654 https://arxiv.org/abs/1904.00625

These and other models are implemented here: https://github.com/black0017/MedicalZooPytorch

[FabianSchuetze]

Could we add SegNext?

Code: https://github.com/visual-attention-network/segnext paper: https://arxiv.org/abs/2209.08575

The code already uses mmsegmentation.

[joihn]

Thanks for this repo, Any plan to support RtFormer ? https://arxiv.org/abs/2210.07124