shicai / DenseNet-Caffe

DenseNet Caffe Models, converted from https://github.com/liuzhuang13/DenseNet
357 stars 273 forks source link

About some computing problem #21

Closed dlyldxwl closed 6 years ago

dlyldxwl commented 6 years ago

您好,有个问题想请教一下。在每个Dense Block中,每个3332的conv前面都接了11128的conv,这样接比不接11128的计算量还要大,因为channel数目增多了,那为什么要这么做呢?

dlyldxwl commented 6 years ago

突然知道了,是因为后面的3*3conv的输入channel远高于128,我只关注了前几层,它们的channel还不到128,所以看起来是增大了计算量,其实对于后面的卷积层来说,大大减少了计算量。