Indian Currency Recognition from Live Video Using Deep Learning

Abstract

Foreign and Visually disable people in India often find difficulties in recognizing different currency notes. Even if some time it is also difficult for Indian healthy people to identify same amount of currency notes with different-new designs. Human eye has also some limitation so some time fake currency not identifiable by them. In this paper using deep learning technique, detection model trained with dataset and tested it with different Indian currency with good accuracy.

VirtualEye: Android Application for the Visually Impaired

“…The authors of this paper say that their approach requires very little data preprocessing and will perform great even if the input images have some disturbances or if the images are unclear. According to [7] paper, an android application was developed by the authors especially for blind people so that they can easily know the denominations of the Indian Banknotes. The authors of this paper have implemented a basic Deep learning model, which scans an image from your smartphone's camera and then gives an output based on some probabilities in the form of voice so that a blind person can hear it easily.…”

“…We would also be making a system where the user has the option to get all outputs in Hindi as well because it should not happen that due to the language barrier, the user is unable to use our application. After reviewing [5], [6], [7] papers we established that the main problem of all the abovementioned approaches is that they employ conventional pre-trained models, such as VGG16/19, AlexNet, Resnet v2, MobileNet, etc., which require a large number of annotated data. The datasets used in the papers do not have all the denominations of currency accepted in India.…”