2021 1st International Conference on Emerging Smart Technologies and Applications (eSmarTA) | 2021
Yemeni Banknote Recognition Model based on Convolution Neural Networks
Abstract
Differentiating between paper currencies with different designs is a challenging task for visually impaired individuals and automated banking machines. Safe and accurate paper currency recognition systems are highly required, because of the wide use of Automated Teller Machines (ATMs), foreign exchange, automatic selling of goods, and automated banking services. With the advances of pattern recognition techniques, many real-life problems have been resolved. This paper presents a robust method to recognize various Yemeni paper currency using pre-trained models. To perform effective recognition processes, deep learning approach is used. Three pre-trained models are implemented which are AlexNet, DenseNet121, and ResNet50. The obtained results of the proposed method achieved high validation accuracy. This robust model reaches a value of 96.99%, 96.83%, and 99.04% in terms of validation accuracy using ResNet-50, DenseNet121, and AlexNet respectively.