DEEP LEARNING FOR GENUINE NAIRA BANKNOTES

Abstract

Over the past few decades, as a result of great technological advancement, currency counterfeiting problems have become more and more serious across the world. Counterfeit notes are currently one of the biggest problems occurring in cash transactions in Nigeria. Therefore, issue of efficient differentiation of counterfeit bank notes from genuine ones via automatic machines has become important. In this work, it is our goal to develop a system that would help to detect counterfeit banknotes and contribute to curbing the menace of currency counterfeiting. To achieve this, we applied a deep learning approach using Faster Region Recurrent Neural Network (FRCNN) to develop a naira detection model in Google Colab. The model is designed to be implemented in a mobile application for fake naira currency recognition called NairaReal. The system was tested with four (4) higher denominations of ₦100, ₦200, ₦500 and ₦1000 currency notes. The result of the overall system returned is an accuracy of 99% for genuine ₦1000 note and fake ₦500 note; and 98% accuracy on both fake ₦1000 and ₦200 notes. The system is relevant to financial institutions, business owners and all citizens who deal with currency transactions on a daily basis.