Refining activation downsampling with SoftPool

Alexandros Stergiou, Ronald Poppe, Grigorios Kalliatakis

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

179 Citations (Scopus)

Abstract

Convolutional Neural Networks (CNNs) use pooling to decrease the size of activation maps. This process is crucial to increase the receptive fields and to reduce computational requirements of subsequent convolutions. An important feature of the pooling operation is the minimization of information loss, with respect to the initial activation maps, without a significant impact on the computation and memory overhead. To meet these requirements, we propose SoftPool: a fast and efficient method for exponentially weighted activation downsampling. Through experiments across a range of architectures and pooling methods, we demonstrate that SoftPool can retain more information in the reduced activation maps. This refined downsampling leads to improvements in a CNN's classification accuracy. Experiments with pooling layer substitutions on ImageNet1K show an increase in accuracy over both original architectures and other pooling methods. We also test SoftPool on video datasets for action recognition. Again, through the direct replacement of pooling layers, we observe consistent performance improvements while computational loads and memory requirements remain limited.

Original languageEnglish
Title of host publication2021 IEEE/CVF International Conference on Computer Vision
Subtitle of host publicationICCV 2021
Pages10337-10346
Number of pages10
DOIs
Publication statusPublished - 2021
Externally publishedYes
Event18th IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2021 - Virtual, Online, Canada
Duration: 11 Oct 202117 Oct 2021
Conference number: 18

Publication series

NameIEEE International Conference on Computer Vision. Proceedings
PublisherIEEE
ISSN (Print)1550-5499
ISSN (Electronic)2380-7504

Conference

Conference18th IEEE/CVF International Conference on Computer Vision Workshops, ICCVW 2021
Abbreviated titleICCVW 2021
Country/TerritoryCanada
CityVirtual, Online
Period11/10/2117/10/21

Keywords

  • n/a OA procedure

Fingerprint

Dive into the research topics of 'Refining activation downsampling with SoftPool'. Together they form a unique fingerprint.

Cite this