Détection de logos pour l'annotation d'images de presse

Pierre Tirilly 1 Vincent Claveau 1, * Patrick Gros 1
* Corresponding author
1 TEXMEX - Multimedia content-based indexing
IRISA - Institut de Recherche en Informatique et Systèmes Aléatoires, Inria Rennes – Bretagne Atlantique
Abstract : In this paper, we propose a new method to annotate news images. To avoid the semantic gap problem due to the use of low-level visual features, we associate high-level visual features (presence of logos and panels) and high-level textual features (a subset of named entities). In one part, we propose a logo and panel detector based on the bag of visual words model as it was proposed by Sivic and Zisserman [1]. The descriptors provided by this model are well suited to the detection of this kind of objects. Our detector requires only a simple learning stage (training data obtained easily, little computation time) and detects logos quickly with a good precision. We evaluate the detector on 413 images from a news corpus. In a second part, we present our annotation method. We associate images that contain logos or panels and some named entities extracted from the news text coming with the images. This annotation method is very fast and is fitted to large-scale applications. We evaluate it on a news corpus that contains more than 40,000 images.
Document type :
Conference papers
Complete list of metadatas

Cited literature [21 references]  Display  Hide  Download

https://hal.inria.fr/inria-00561796
Contributor : Patrick Gros <>
Submitted on : Tuesday, February 1, 2011 - 7:18:40 PM
Last modification on : Friday, November 16, 2018 - 1:21:59 AM
Long-term archiving on : Tuesday, November 6, 2012 - 1:10:32 PM

File

Tirilly-et-al-RFIA2010.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00561796, version 1

Citation

Pierre Tirilly, Vincent Claveau, Patrick Gros. Détection de logos pour l'annotation d'images de presse. Congrès francophone AFRIF-AFIA de reconnaissance de formes et d'intelligence artificielle, RFIA'10, AFRIF - AFIA, Jan 2010, Caen, France. ⟨inria-00561796⟩

Share

Metrics

Record views

406

Files downloads

270