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

Link to Paper:

“Squeeze-and-Excitation Networks” - 2017

arxiv.org

Table of Contents

1. Introduction

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

Jie Hu | Gang Sun | Li Shen | Enhua Wu | Samuel Albanie

Contributions

• 1st Place in the last ImageNet Large Scale Visual Recognition Challenge(LSVRC) in 2017
• Introduced new architectural unit, Squeeze-and-Excitation (SE) block, to
  • improve the quality of representations produced by a network by explicitly modelling the interdependencies between the channels of its convolutional features
  • perform feature recalibration, through which it can learn to use global information to selectively emphasise informative features and suppress less useful ones.

2. Architecture