Fusion multimodale image/texte par réseaux de neurones profonds pour la classification de documents imprimés

Abstract : Document classification is an important task in the analysis and processing of digital collections as it is mainly used for input pipeline of such systems. To extract features allowing algorithms to categorize the elements, text and pictures are used. We present in this paper different approaches for document classification using textual datas and pictures, as well as a classification model using both of this datas in single model of convolution neural network.
Complete list of metadatas

Cited literature [17 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01905242
Contributor : Thibault Magallon <>
Submitted on : Thursday, October 25, 2018 - 4:17:28 PM
Last modification on : Tuesday, December 18, 2018 - 8:04:08 AM
Long-term archiving on : Saturday, January 26, 2019 - 3:40:32 PM

File

CORIA_VFinal.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01905242, version 1

Collections

Citation

Thibault Magallon, Frédéric Béchet, Benoit Favre. Fusion multimodale image/texte par réseaux de neurones profonds pour la classification de documents imprimés. 15e Conférence en Recherche d’Information et Applications (CORIA), May 2018, Rennes, France. ⟨hal-01905242⟩

Share

Metrics

Record views

80

Files downloads

80