Skip to content
ml-recall

#depthwise-separable-convolution

2 questions from Convolutional Neural Networks, Landmark Architectures

Timed test

2 questions · 2 min · score and explanations at the end

Take a test
Q 1 / 2Score 0 / 0

Question 1: Short answer

mediumAI-checkedConvolutional Neural Networks

A convolution maps 64 input channels to 2 output channels with 3×3 kernels (no biases). A depthwise separable version uses one 3×3 filter per input channel (depthwise step), followed by a 1×1 convolution from 64 to 2 channels (pointwise step). How many fewer weights does the depthwise separable version need?

an integer