search

weiliu89 / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
4.77k stars 1.68k forks source link

Resnet-101 for SSD #758

Open caiqi opened 7 years ago

caiqi commented 7 years ago

According to the DSSD paper, "Following SSD, and fitting the Residual architecture,...." the added extra layers are Residual blocks, however, in this repo, it's just two conv layers. Is there any benefit of using Residual blocks? And for the residual blocks, is the paper using kernel size = 33 stride = 2 or kernel size=11 and stride =2? I can't find details in the paper. Besides, do you have any plan of releasing the code of DSSD? Thanks!

CrazySssst commented 7 years ago

the code of DSSD is here: https://github.com/chengyangfu/caffe/tree/dssd

gachiemchiep commented 6 years ago

@CrazySssst Thank you man.