Kronecker decomposition for image classification

Sabrina Fontanella*, Antonio J. Rodríguez-Sánchez, Justus Piater, Sandor Szedmak

*Corresponding author for this work

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

Abstract

We propose an image decomposition technique that captures the structure of a scene. An image is decomposed into a matrix that represents the adjacency between the elements of the image and their distance. Images decomposed this way are then classified using a maximum margin regression (MMR) approach where the normal vector of the separating hyperplane maps the input feature vectors into the outputs vectors. Multiclass and multilabel classification are native to MMR, unlike other more classical maximum margin approaches, like SVM. We have tested our approach with the ImageCLEF 2015 multi-label classification task, Pascal VOC and Flickr dataset.

Original languageEnglish
Title of host publicationExperimental IR Meets Multilinguality, Multimodality, and Interaction - 7th International Conference of the CLEF Association, CLEF 2016, Proceedings
PublisherSpringer
Pages137-149
Number of pages13
Volume9822
ISBN (Print)9783319445632
DOIs
Publication statusPublished - 2016
MoE publication typeA4 Conference publication
EventInternational Conference of the CLEF Association - Evora, Portugal
Duration: 5 Sept 20168 Sept 2016
Conference number: 7

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9822
ISSN (Print)03029743
ISSN (Electronic)16113349

Conference

ConferenceInternational Conference of the CLEF Association
Abbreviated titleCLEF
Country/TerritoryPortugal
CityEvora
Period05/09/201608/09/2016

Keywords

  • ImageCLEF
  • Kronecker decomposition
  • Maximum margin
  • Medical images
  • MMR
  • Multi-label classification
  • SVM

Fingerprint

Dive into the research topics of 'Kronecker decomposition for image classification'. Together they form a unique fingerprint.

Cite this