Identifikasi Nominal Mata Uang Rupiah Bagi Penyandang Tunanetra Dengan Algoritma Convolutional Neural Network Berbasis Android

Abstract

What frequently occurs in today's life is the process of economic transactions, particularly during buying and selling. When engaging in these transactions, people always utilize money as the payment method. This implies that money serves a rather crucial function during transactions. However, for visually impaired individuals, the transaction process becomes problematic because they face difficulties in recognizing the denominations on banknotes. Deep Learning is a burgeoning field within Machine Learning that has shown significant development. Among data practitioners, Deep Learning is widely popular due to its exceptional capabilities in computer vision. One notable application is object classification in images. By implementing one of the Machine Learning methods that can be employed to classify Indonesian rupiah currency, Convolutional Neural Network (CNN) can be utilized. Based on the constructed architecture, the Convolutional Neural Network (CNN) method encompasses various types, with VGG-16 being one of them. In this study, 7 classes of Indonesian rupiah currency images were employed, each consisting of 190 images. The obtained accuracy rate is 83% when utilizing the VGG-16 architecture.