An Automated System for Indian Currency Classification and Detection using CNN

Abstract

The visibly disabled frequently experience difficulties with daily tasks that a typical person would take for granted, such as handling financial transactions. Since India’s demonetization took effect, accepting money has become a difficult task. Due to the similar dimensions of new cash banknotes and the fact that some old money banknotes are still in use, India now has two banknotes for every category. Due to the current situation, it is extremely difficult to identify banknotes for those who appear to be weak. The sight and brain are gifts to humans. Detecting things with the same qualities is practically impossible for persons who are sight impaired. In keeping with this, we suggest an automated system that would enable those who are visually impaired to recognize currency through a sound notification from a variety of applications. Therefore, in this quest, we help them locate the currency notes. In this study, we apply different convolution neural network (CNN) models to datasets of Indian banknotes in order to extract deep features and recognize different currencies. To train, verify, and test the CNN model, we can produce a fresh dataset of Indian banknotes. The proposed model may be created with TensorFlow, enhanced by choosing the best hyper parameter value, and evaluated against well-established CNN architectures using transfer learning.

Keywords

• convolutional neural networks (cnns)
• deep learning
• vgg16
• alexnet
• mobilenet
• tensorflow
• accuracy