Handwritten text recognition: from isolated text lines to whole documents - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

Handwritten text recognition: from isolated text lines to whole documents

Résumé

The handwriting recognition task is largely dominated by deep neural networks. However, it remains challenging for these advanced computer vision systems. Recently, the models have become more sophisticated, moving from line-level recognition to paragraph-level and even page-level recognition. In this paper, we will study those advances and the constraints that come with them, mainly focusing on two models we proposed: the Simple Predict & Align Network and the Vertical Attention Network. Both handle paragraph images, and we outperformed the state of the art on three datasets: RIMES, IAM and READ 2016
Fichier principal
Vignette du fichier
2021_ORASIS.pdf (2.61 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03339648 , version 1 (09-09-2021)

Identifiants

  • HAL Id : hal-03339648 , version 1

Citer

Denis Coquenet, Clément Chatelain, Thierry Paquet. Handwritten text recognition: from isolated text lines to whole documents. ORASIS 2021, Centre National de la Recherche Scientifique [CNRS], Sep 2021, Saint Ferréol, France. ⟨hal-03339648⟩
243 Consultations
64 Téléchargements

Partager

Gmail Facebook X LinkedIn More