TUTCRIS - Tampereen teknillinen yliopisto

TUTCRIS

Graph-regularized Multi-class Support Vector Machines for Face and Action Recognition

Tutkimustuotosvertaisarvioitu

Yksityiskohdat

AlkuperäiskieliEnglanti
Otsikko2016 24th European Signal Processing Conference (EUSIPCO)
KustantajaIEEE
Sivut96-100
Sivumäärä5
ISBN (elektroninen)978-0-9928-6265-7
TilaJulkaistu - 2016
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisussa
TapahtumaEUROPEAN SIGNAL PROCESSING CONFERENCE -
Kesto: 1 tammikuuta 1900 → …

Julkaisusarja

Nimi
ISSN (elektroninen)2076-1465

Conference

ConferenceEUROPEAN SIGNAL PROCESSING CONFERENCE
Ajanjakso1/01/00 → …

Tiivistelmä

In this paper, we formulate a variant of the Support Vector Machine classifier that exploits graph-based discrimination criteria within a multi-class optimization process. We employ two kNN graphs in order to describe intra-class and betweenclass data relationships. These graph structures are combined in order to form a regularizer which is used in order to regularize the multi-class SVM optimization problem. The derived multiclass classifier is compared with the standard SVM classifier and SVM formulations exploiting geometric class information on six publicly available databases designed for human action recognition in the wild and facial image classification problems, where its effectiveness is shown.