We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
We adopt convolution depthwise layer in deploy_xception-dw.prototxt for speeding up.