$\color{black}\rule{365px}{3px}$

Link to Paper:

“MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications” - 2017

arxiv.org

Table of Contents

1. Introduction

$\color{black}\rule{365px}{3px}$

Motivation

“advances to improve accuracy are not necessarily making networks more efficient with respect to size and speed. In many real world applications such as robotics, self-driving car and augmented reality, the recognition tasks need to be carried out in a timely fashion on a computationally limited platform.”

Contributions

• Use of Depthwise Separable Convolution to replace standard convolutions:

  A standard convolution operation combines spatial filtering and feature channel mixing in one step. Depthwise separable convolutions break this process into two smaller, more efficient steps:

  • Depthwise Convolution: Applies a single convolutional filter to each input channel (spatial filtering is done channel-wise).
  • Pointwise Convolution: Uses 1x1 convolutions to mix the output channels generated by the depthwise convolution.