LikeLy-Journey / SegmenTron

Support PointRend, Fast_SCNN, HRNet, Deeplabv3_plus(xception, resnet, mobilenet), ContextNet, FPENet, DABNet, EdaNet, ENet, Espnetv2, RefineNet, UNet, DANet, HRNet, DFANet, HardNet, LedNet, OCNet, EncNet, DuNet, CGNet, CCNet, BiSeNet, PSPNet, ICNet, FCN, deeplab)
Apache License 2.0
692 stars 162 forks source link

HardNet or FCHardNet? #48

Open InternetMaster1 opened 4 years ago

InternetMaster1 commented 4 years ago

Which is the HardNet architecture implemeted in this library?

Is it FC-HarDNet-70 from https://github.com/PingoLH/FCHarDNet

LikeLy-Journey commented 4 years ago

yes

InternetMaster1 commented 4 years ago

@LikeLy-Journey Thanks!

1) License of HardNet?

The https://github.com/PingoLH/FCHarDNet library doesn't mention a license for their technology. Is that the official library for HarDNet architecture?

Your library has a MIT license. So is it okay to use your HardNet implementation for commercial purposes?

2) HRNet

Any plans to add HRNetV2-W18-Small-v2 to your library? It has an impressive mIOU of 76.2

I wonder if they have a "HRNet + OCR + SegFix" option available for HRNetV2-W18-Small-v2? That seems to increase accuracy

3) SOTA for lightweight segmentation?

I am looking to implement high-quality semantic segmentation on a mobile device for accurate human segmentation for still images (i.e. non-realtime).

I have been researching a lot of lightweight architectures/backbones for this purpose (HardNet, MobileNetV2, MobileNetV2plus, ShelfNet, FasterSeg, BiseNetV2, SINet, PortraitNet, etc)

Could HRNetV2-W18-Small-v2 be considered SOTA for this task? It achieves mIOU of 76.2% on Cityscapes.

The mIOU of U-HarDNet-70 is 75.9% & it is at the top of paperswithcode SOTA list for real-time segmentation.

4) Is the CityScapes leaderboard a good place to look for my task at hand? I am looking for human/person segmentation. Should I rather be looking for leaderboards of Coco or PASCAL? Or it doesn't really matter?

Thanks!