On Multi-class Currency Classification Using Convolutional Neural Networks and Cloud Computing Systems for the Blind

Abstract

Identifying a currency is the easiest of task for a person. But for a blind human, it is not always the same. Although the currencies come in all shapes and sizes, it is always possible for a blind person to get confused with the currency value. We analyze the potential usage of cloud computing systems, image processing techniques along with deep learning approaches for identifying the currency based on its denomination which can be accessed by the blind via a mobile app. We examine different approaches including public cloud and private cloud systems for training and deploying the deep learning techniques. We achieve an accuracy of up to 98% in our proposed models. We propose models which can be used under a variety of situations including cost-effective approaches, efficient approaches that suite reduced local storage, etc.