Closed extendedcaffe closed 7 years ago
Hi dear, thanks for sharing, your 3D Caffe and VNet is very useful and educational. I'm wondering why you use convolution layer in your prototxt to do pooling? are there any accuracy issue/special consideration? thanks.
@extendedcaffe
Strided covolution turned better approach than pooling .
Read the paper in citation "striving for simplicity the all convolutional net"
you will get your answer
Hi dear, thanks for sharing, your 3D Caffe and VNet is very useful and educational. I'm wondering why you use convolution layer in your prototxt to do pooling? are there any accuracy issue/special consideration? thanks.