Currency Detection for Visually Impaired People using Edge Devices
Ujjwal Kadam, Arvind Meena, Chaudhary Abuzar, Ujjwal, Gaurav Singhal, Divya Srivastava · 2023
Visual impairment is a global issue that affects a significant number of individuals. In India, there are nearly 4.95 million blind people and 70 million vision-impaired people and many of them have difficulty reading currency notes and struggle with performing their everyday transactions. There is a need to develop a device that can help them read currency notes and perform their everyday transactions smoothly. In this paper, we have a solution for the same. An edge device, embedded in a cane stick, that can detect the currency notes using a CNN model and provide audio output. We have improved upon an existing INR (Indian National Rupees) currency dataset by adding additional images to it. We trained a CNN model that can classify INR currency images and deployed it on a Raspberry Pi. The TensorFlow model that we have deployed boasts better accuracy compared to other existing models and has been trained over the largest existing INR dataset.