DEEP LEARNING APPROACH FOR INDIAN CURRENCY CLASSIFICATION
Suyash Mahesh Bahrani · International Journal of Engineering Applied Sciences and Technology · 2020
Currency is an unavoidable part of our day-today life.Despite the rapidly expanding utilization of master cards and additional electronic payment categories, money is considerably utilized for everyday exchanges because of its comfort.The current day monetary self-service gives birth to currency recognition, which plays a vital role in the automated banking procedure.Therefore we propose a novel method for currency recognition that identifies Indian currency in different views on the scale.It is straightforward for a typical human being to comprehend and recognize any banknote easily, but it is undoubtedly troublesome for anyone with a visually impaired or blind individual to accomplish a similar task.Banknotes commonly have unique designs according to the denomination and can be sorted with surplus human errors in the bank.These errors lead to difficulties in evaluating and recognition.If computers or mobile apps recognize currency, it will immensely boost the precision of recognition and ameliorate people's workload efficiently.As money has a significant role in daily life for any business transactions, real-time detection and recognition of banknotes become necessary for a person, especially blind or visually impaired, or for a system that sorts the data.The model which we worked on essentially classifies the currency note into distinct denominations like Rs10, Rs50, Rs100, Rs500, Rs2000.The currency will be recognized and classified by using image processing techniques, deep learning techniques.We have implemented transfer learning theory, a deep learning domain where we will reuse the weights in one or more layers from an already trained model into a new model by either maintaining the weights fixed, fine-tuning them, or adapting the weights completely when training the model.We aim to enhance techniques that have been missing in most contemporary works that have been done so far.Therefore our proposed currency recognition system can be efficiently run on the web application/mobile app, where a user would upload an image of the currency note, and it will deliver an audio output and a text output.