deep-high-resolution-net.pytorch icon indicating copy to clipboard operation
deep-high-resolution-net.pytorch copied to clipboard

[Network structure] Why hrnet use basicblock in stage 2/3/4

Open HeathHose opened this issue 6 years ago • 3 comments

Thanks for your works! In Deep residual network, resnet34 use basicblock and resnet50/101 use bottleneck for practice performance. So what about hrnet? If hrnet replace basicblock by bottleneck in stage 2/3/4, the accuracy will be dropped?

looking forward to the reply!

HeathHose avatar Sep 10 '19 12:09 HeathHose

I also want to know the result.

gaoyao123 avatar Sep 25 '19 13:09 gaoyao123

Joining the question.

peternaf avatar May 20 '20 10:05 peternaf

Thanks for your works! In Deep residual network, resnet34 use basicblock and resnet50/101 use bottleneck for practice performance. So what about hrnet? If hrnet replace basicblock by bottleneck in stage 2/3/4, the accuracy will be dropped?

looking forward to the reply!

I found that when you use bottleneck block to replace the original basic block, the number of parameters will increase dramatically...It's counter-intuitive.

seekFire avatar Nov 03 '21 13:11 seekFire