Skip to main navigation Skip to search Skip to main content

An encoder generative adversarial network for multi-modality image recognition

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

Abstract

This paper is concerned with the multi-modality image recognition which is a crucial technique used in industrial applications. The paper proposes a novel algorithm based on the deep generative adversarial network to learn a common feature space between different modalities. These abstract features are robust to the modality discrepancy and can be used to train a cross-modality classifier which will achieve excellent performance on all modalities. The comparative experiments on standard mul-ti-modality image recognition benchmark are employed to validate the effectiveness of our proposed algorithm. The results demonstrate that the proposed network is efficient to deal with the multi-modality recognition challenge, especially improve the performance on the modalities with limited training samples.

Original languageEnglish
Title of host publicationProceedings
Subtitle of host publicationIECON 2018 - 44th Annual Conference of the IEEE Industrial Electronics Society
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages2689-2694
Number of pages6
ISBN (Electronic)9781509066841
DOIs
StatePublished - 26 Dec 2018
Event44th Annual Conference of the IEEE Industrial Electronics Society, IECON 2018 - Washington, United States
Duration: 20 Oct 201823 Oct 2018

Publication series

NameProceedings: IECON 2018 - 44th Annual Conference of the IEEE Industrial Electronics Society

Conference

Conference44th Annual Conference of the IEEE Industrial Electronics Society, IECON 2018
Country/TerritoryUnited States
CityWashington
Period20/10/1823/10/18

Keywords

  • Deep learning
  • Generative adversarial network
  • Image recognition
  • Multi-modality

Fingerprint

Dive into the research topics of 'An encoder generative adversarial network for multi-modality image recognition'. Together they form a unique fingerprint.

Cite this