DEVELOPMENT OF A MOBILE APP FOR NAIRA-CURRENCY IDENTIFICATION USING DEEP LEARNING CLASSIFICATION

ABSTRACT

Currency recognition is vital for banking automation, accessibility support for visually impaired individuals, and counterfeit detection. In Nigeria, the need for a robust, locally optimized mobile-based currency detection system is particularly significant due to the increasing use of smartphones and the threat of counterfeit currency. This research is focused on developing and implementing a deep learning-based mobile phone application with the capability to recognize and verify Nigerian currency (N1000 and N500 denominations, including counterfeits) using Convolutional Neural Networks (CNN). The methodology involved collecting and preprocessing a custom dataset under various real-world environments, training a CNN model using TensorFlow and Keras, and deploying the optimized model to Android devices via TensorFlow Lite. Results showed a validation accuracy of 100% and real-world mobile test accuracy of approximately 96%, with high great sensitivity, specificity, and precision for all classes. The model is effectively distinguished between genuine and counterfeit notes, even under variable lighting conditions and orientations. In conclusion, the proposed mobile application demonstrates significant potential in supporting inclusive financial interactions and secure transactions in Nigeria. Future work may explore extending the model to more denominations, cross-platform deployment, and real-time performance optimization under extreme conditions.

Keywords: TensorFlow, Naira, Image processing, Convolutional Neural Network, Android

Keywords: Naira currency recognition, Convolutional Neural Network (CNN), Mobile application, Counterfeit detection, TensorFlow Lite Image classification