Facing scalability: Naming faces in an online social network

Ronald Walter Poppe

Research output: Contribution to journalArticleAcademicpeer-review

8 Citations (Scopus)

Abstract

Automatically naming faces in online social networks enables us to search for photos and build user face models. We consider two common weakly supervised settings where: (1) users are linked to photos, not to faces and (2) photos are not labeled but part of a user's album. The focus is on algorithms that scale up to an entire online social network. We extensively evaluate different graph-based strategies to label faces in both settings and consider dependencies. We achieve results on a par with a recent multi-person approach, but with 60 times less computation time on a set of 300K weakly labeled faces and 1.4 M faces in user albums. A subset of the faces can be labeled with a speed-up of over three orders of magnitude.
Original languageUndefined
Pages (from-to)2335-2347
Number of pages13
JournalPattern recognition
Volume45
Issue number6
DOIs
Publication statusPublished - Jun 2012

Keywords

  • HMI-CI: Computational Intelligence
  • HMI-MR: MULTIMEDIA RETRIEVAL
  • Weak labeling
  • Online Social Network
  • EWI-21519
  • METIS-285135
  • IR-79722
  • Graph-Based
  • Face naming
  • Scalability

Cite this

Poppe, Ronald Walter. / Facing scalability: Naming faces in an online social network. In: Pattern recognition. 2012 ; Vol. 45, No. 6. pp. 2335-2347.
@article{cca3fdf2785d43a7a1f1f30d67963f1e,
title = "Facing scalability: Naming faces in an online social network",
abstract = "Automatically naming faces in online social networks enables us to search for photos and build user face models. We consider two common weakly supervised settings where: (1) users are linked to photos, not to faces and (2) photos are not labeled but part of a user's album. The focus is on algorithms that scale up to an entire online social network. We extensively evaluate different graph-based strategies to label faces in both settings and consider dependencies. We achieve results on a par with a recent multi-person approach, but with 60 times less computation time on a set of 300K weakly labeled faces and 1.4 M faces in user albums. A subset of the faces can be labeled with a speed-up of over three orders of magnitude.",
keywords = "HMI-CI: Computational Intelligence, HMI-MR: MULTIMEDIA RETRIEVAL, Weak labeling, Online Social Network, EWI-21519, METIS-285135, IR-79722, Graph-Based, Face naming, Scalability",
author = "Poppe, {Ronald Walter}",
note = "eemcs-eprint-21519",
year = "2012",
month = "6",
doi = "10.1016/j.patcog.2011.12.018",
language = "Undefined",
volume = "45",
pages = "2335--2347",
journal = "Pattern recognition",
issn = "0031-3203",
publisher = "Elsevier",
number = "6",

}

Facing scalability: Naming faces in an online social network. / Poppe, Ronald Walter.

In: Pattern recognition, Vol. 45, No. 6, 06.2012, p. 2335-2347.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Facing scalability: Naming faces in an online social network

AU - Poppe, Ronald Walter

N1 - eemcs-eprint-21519

PY - 2012/6

Y1 - 2012/6

N2 - Automatically naming faces in online social networks enables us to search for photos and build user face models. We consider two common weakly supervised settings where: (1) users are linked to photos, not to faces and (2) photos are not labeled but part of a user's album. The focus is on algorithms that scale up to an entire online social network. We extensively evaluate different graph-based strategies to label faces in both settings and consider dependencies. We achieve results on a par with a recent multi-person approach, but with 60 times less computation time on a set of 300K weakly labeled faces and 1.4 M faces in user albums. A subset of the faces can be labeled with a speed-up of over three orders of magnitude.

AB - Automatically naming faces in online social networks enables us to search for photos and build user face models. We consider two common weakly supervised settings where: (1) users are linked to photos, not to faces and (2) photos are not labeled but part of a user's album. The focus is on algorithms that scale up to an entire online social network. We extensively evaluate different graph-based strategies to label faces in both settings and consider dependencies. We achieve results on a par with a recent multi-person approach, but with 60 times less computation time on a set of 300K weakly labeled faces and 1.4 M faces in user albums. A subset of the faces can be labeled with a speed-up of over three orders of magnitude.

KW - HMI-CI: Computational Intelligence

KW - HMI-MR: MULTIMEDIA RETRIEVAL

KW - Weak labeling

KW - Online Social Network

KW - EWI-21519

KW - METIS-285135

KW - IR-79722

KW - Graph-Based

KW - Face naming

KW - Scalability

U2 - 10.1016/j.patcog.2011.12.018

DO - 10.1016/j.patcog.2011.12.018

M3 - Article

VL - 45

SP - 2335

EP - 2347

JO - Pattern recognition

JF - Pattern recognition

SN - 0031-3203

IS - 6

ER -