Questions about Channel-wise Separable Convolutions about region-based-non-local-network HOT 3 OPEN

Hi,

Thank you for interesting my work.

For the implementation of Fθ, I did use channel-wise(depth-wise) convolution, which mean the numbe of groups is equal to the channel number. However, in Pytorch, setting the groups parameter as the channel number for a Conv would cuase low efficency. I guess Pytorch1.3 implemented it with a loop operation. So I code the DepthwiseConv3d myself by setting masks for some channels.

I have tested that in Pytorch-1.6 or higher versions, they have make the Convolution with multiple goups more efficient. But considering that the trained models still use Pytorch1.3, so I did not change it.

from region-based-non-local-network.

Hi,

Thanks for the reply!
It solved all my problems.

from region-based-non-local-network.

Hi,

Thanks for the reply!
It solved all my problems.

from region-based-non-local-network.