Delving deep into spatial pooling for squeeze-and-excitation networks

作者：

Highlights：

• We revisit the squeeze operation in SENets, and shed lights on why and how to embed rich (both global and local) spatial information into the excitation module to improve accuracy.

• We propose an integrated two-stage spatial pooling method with two efficient implementation approaches for rich descriptor extraction.

• We conduct extensive experiments to verify convincing improvements over SENets and their extension on various fundamental computer vision tasks.

摘要

•We revisit the squeeze operation in SENets, and shed lights on why and how to embed rich (both global and local) spatial information into the excitation module to improve accuracy.•We propose an integrated two-stage spatial pooling method with two efficient implementation approaches for rich descriptor extraction.•We conduct extensive experiments to verify convincing improvements over SENets and their extension on various fundamental computer vision tasks.

论文关键词：Convolutional neural networks,Squeeze-and-excitation,Spatial pooling,Base model

论文评审过程：Received 4 March 2020, Revised 10 February 2021, Accepted 4 July 2021, Available online 15 July 2021, Version of Record 29 July 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108159