search

VDIGPKU / CBNet_caffe

Composite Backbone Network (AAAI20)
Apache License 2.0
408 stars 78 forks source link

Where is the BN layer before upsampling? #20

Closed panjianning closed 4 years ago

panjianning commented 4 years ago

The paper said

the composite connection, which consists of a 1×1 convolutional layer and batch normalization layer to reduce the channels and an upsample operation

But I can't find BN before upsampling here

image

panjianning commented 4 years ago

I believe batch norm is not used here since you add bias term in Conv and Conv with bias = Conv without bias + not-collecting-runining-stat BN.