Iles d Imesli
Volume 13, Numéro 2, Pages 21-27
2021-12-31

Handwritten Characters Recognition For Amazigh Script

Authors : Sadou Malika .

Abstract

Amazigh natural language processing faces several challenges, including the ununified Amazigh graphical system and the unavailability of large public databases. In this paper, we elaborate a deep learning architecture that can be effectively applied to recognizing Amazigh handwritten characters. It concerns the implantation of a CNN network, which is applied and tested on the BERBER-MNIST dataset. The experimental results showed that the proposed CNN achieved a very interesting recognition rate.

Keywords

BERBER-MNIST ; CNN ; Characters Recongnition ; Amazigh Script ; Deep learning.