Money Identifier: An Android-Based Application For Visually Impaired People Using Object Detection

Abstract

A normal person can readily recognize and distinguish any banknote, but a visually impaired person finds the task much more difficult as they either have partial or complete vision loss. They face several obstacles when going about their daily lives. They have many issues in terms of monetary transactions as they cannot recognize the currencies due to the similarity of paper texture and size between different categories. To address this problem, the proponents implemented image processing techniques that will assist visually impaired people in detecting and identifying money using object detection. The objective of this application is to make them feel secure and confident in their financial decisions. It will be implemented as an Android-based money detection app. Also, the researchers utilized AGILE and TensorFlow's Convolutional Neural Network algorithm to build a model. They collected multiple images of Philippines notes captured in different light conditions and angles to achieve efficient results in the Money Identifier Application. Furthermore, the proponents have created a survey questionnaire for blind persons to evaluate the produced application. The Eyessential program can recognize Philippine money with a mean of 4.91, a 98 percent accuracy, and a run time of 3 to 4 seconds. The software is functional, dependable, helpful, efficient, maintainable, and portable based on the results. The developers suggest submitting the Eyessential Android app to Google Play to allow individuals to install the software from anywhere. Also, Google Play services work with Google Assistant, useful for blind people who want to utilize the app. The developers propose adding a flashlight option to make the program more user-friendly. To detect denominations precisely, the developers suggest employing substantial data sets