Mobile resnest

[Jerryzcn]

1. resnest14
2. resnest26

[zhanghang1989]

👍 will merge after uploading the weights.

[InternetMaster1]

@Jerryzcn

Have you been able to run ResNeST on mobile? What is the accuracy you are getting for 14 & 26?

[zhanghang1989]

@Jerryzcn

Have you been able to run ResNeST on mobile? What is the accuracy you are getting for 14 & 26?

The results are reported https://gluon-cv.mxnet.io/model_zoo/classification.html#resnest

Model	top-1	top-5
ResNeSt14	75.75	92.70
ResNeSt26	78.68	94.38

[InternetMaster1]

Sorry, I meant to ask about semantic segmentation.. I am looking for a high accuracy architecture/backbone for mobile (mainly for human segmentation). ResNest mIOU looked good but the param size is large. Any recommendations?

…

On Wed, 6 May, 2020, 10:40 PM Hang Zhang, ***@***.***> wrote: @Jerryzcn <https://github.com/Jerryzcn> Have you been able to run ResNeST on mobile? What is the accuracy you are getting for 14 & 26? The results are reported https://gluon-cv.mxnet.io/model_zoo/classification.html#resnest Model top-1 top-5 ResNeSt14 75.75 92.70 ResNeSt26 78.68 94.38 — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#29 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD4AFEB7AW3LVLLD3VACTV3RQGKXVANCNFSM4MOX5XSQ> .

[DKandrew]

@Jerryzcn How do you come up with the name resnest14 and resnest26? I count the number of Conv2d in the resnest26 and they are definitely not equal to 26.

[FrancescoSaverioZuppichini]

@DKandrew That's the number of layers!