Indian Currency Recognition System Using CNN And Comparison With YOLOv5

Abstract

Computer vision is the most anticipated technology of the 21st century. Object detection is the basic functional block in Convolution Neutral Network. The objective of the proposed methodology is to design a system to detect Indian currencies which can help visually impaired people to recognize and read out the value of all possible Indian paper currencies with more than 79.83% accuracy. In this approach, the features of the notes are extracted in separate Red Green Blue (RGB) layers, normalized and quantized into machine readable data that is later trained with Adam optimizer which gives Probabilistic Prediction of each type of currency with a loud audio output. The Convolution Neural Network model is further compared with Yolov5 model which is considered to be the fastest algorithm for object detection. After comparison the accuracy of Convolution Neural Network model was found to be just 1% lesser than that of the YOLOv5 model.