Design & Build Banknote Nominal Identification Tools for Visual Impairment Using Convolutional Neural Network Algorithm and Tensorflow with Android Based

Abstract

The buying and selling transactions that we usually do in our daily lives are difficult for people with visual impairment because it is difficult to distinguish the denomination of rupiah banknotes because of the limitation of sight, so it becomes one of the problems. This research focused on making a voice-based rupiah banknote nominal detection tool, using convolutional neural network algorithms in machine learning as the core of this system. This tool was also equipped with a voice-based android application to monitor the remaining money used when making buying and selling transactions-testing the tool used real test data of 20 images per class, producing an accuracy of 83%, as evidenced by Confusion matrix calculations.