search

HRNet / Lite-HRNet

This is an official pytorch implementation of Lite-HRNet: A Lightweight High-Resolution Network.
Apache License 2.0
820 stars 126 forks source link

Training speed decreases #15

Open huixiancheng opened 3 years ago

huixiancheng commented 3 years ago

I used it on semantic segmentation and found that the loss convergence speed is very very slow(compare with other backbone)

EEWenbinWu commented 3 years ago

I used it on pose estimation and I have the same problem.

huixiancheng commented 3 years ago

maybe it's cause by depthwise convolution

nsapkota417 commented 2 years ago

I used it on semantic segmentation and found that the loss convergence speed is very very slow(compare with other backbone)

Hi, what was your segmentation performance like? Were you trying on Cityscape dataset?

huixiancheng commented 2 years ago

No. Finally I give up.