L’équipe ARMEDIA est heureuse de vous convier à un séminaire présenté par M. Abir Fathallah, PhD student, Telecom SudParis

Quand : Jeudi 27 mai 2021 à 14h30

Pour des raisons sanitaires, les séminaires seront tenus en distanciel sur l’espace suivant : https://webconf.imt.fr/frontend/cat-clh-6al

Title : « Triplet CNN-based Word Spotting of Historical Arabic Documents »

Abstract :
Word Spotting of Historical Arabic Documents has remained a challenging task due to the complexity of documents layout. This paper proposes a novel word spotting system that consists of learning feature representation to describe word images. The objective is to investigate optimal embedding spaces to extract a discriminative word image representation. Our proposed approach consists of two steps : i) construct a CNN-based embedding space with triplet-loss and then ii) match embedding representations using the euclidean distance. For training, the CNN takes as input a set of triplet samples (anchor, positive sample and negative sample). Then, triplet loss serves to create a novel space by minimizing intra-classes distances and maximizing inter-classes distances. The proposed approach is evaluated on the VML-HD dataset and the experiments show its effectiveness compared to the state of the art.