Performance Evaluation of Symbol Recognition

Ernest Valveny Philippe Dosch 1
1 QGAR - Querying Graphics through Analysis and Recognition
INRIA Lorraine, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
Abstract : Symbol recognition is one of the central problems in the field of graphics recognition. Many methods and approaches have been developed in the context of several application domains. In the last years, the need for generic methods, able to perform well on large sets of symbols in different domains, has become clear. Thus, standard evaluation datasets and protocols have to be built up in order to be able to evaluate the performance of all these methods. In this paper we discuss several points which should be taken into account in the design of such evaluation framework, raising a number of open questions for further discussion. These issues were the starting point of the organization of the contest on symbol recognition held during the las Workshop on Graphics Recognition (GREC'03). We also summarize the main features of the dataset and the protocol of evaluation used in the contest, as a first step to define a general evaluation framework, giving answer to these open questions.
Type de document :
Communication dans un congrès
Simone Marinai and Andreas Dengel. 6th IAPR International Workshop on Document Analysis Systems - DAS'04, 2004, Florence, Italie, Springer Verlag, 3163, pp.354-365, 2004, Lecture Notes in Computer Science
Liste complète des métadonnées

https://hal.inria.fr/inria-00100000
Contributeur : Publications Loria <>
Soumis le : mardi 26 septembre 2006 - 10:13:17
Dernière modification le : jeudi 11 janvier 2018 - 06:19:59

Identifiants

  • HAL Id : inria-00100000, version 1

Collections

Citation

Ernest Valveny, Philippe Dosch. Performance Evaluation of Symbol Recognition. Simone Marinai and Andreas Dengel. 6th IAPR International Workshop on Document Analysis Systems - DAS'04, 2004, Florence, Italie, Springer Verlag, 3163, pp.354-365, 2004, Lecture Notes in Computer Science. 〈inria-00100000〉

Partager

Métriques

Consultations de la notice

93