Depthwise Separable Convolution

What is Depthwise Separable Convolution?

An efficient convolution variant that splits the operation into a depthwise spatial filter and a pointwise (1x1) channel mixer, drastically reducing parameters.

Where did the term "Depthwise Separable Convolution" come from?

Core of MobileNet architecture.

How is "Depthwise Separable Convolution" used today?

Enables efficient vision on phones.

Related Terms

• Convolutional Neural Network (CNN)
• Deep Learning