Development and evaluation of a banknote reader for visually impaired people using deep learning-based object detection

Chan, Hong Wai (2023) Development and evaluation of a banknote reader for visually impaired people using deep learning-based object detection. Final Year Project, UTAR.

Preview

PDF Download (3570Kb) | Preview

Abstract

Automated currency recognition systems have numerous applications, including banknote counting machines, currency exchange machines, and systems to assist visually impaired individuals. The inability to differentiate between different currencies can lead to financial exploitation of visually impaired individuals, making the need for a reliable currency recognition system even more demanding. In this project, we propose a mobile system for currency recognition that can recognize the Malaysian banknotes and sen of different denominations using deep learning techniques. This project introduces a system using the YOLOv8 architecture for feature extraction and classification, designed to benefit the visually impaired community. The project is structured into two primary phases: the training phase, which encompasses model training and exportation, and the development phase, including model integration and the deployment of the user-friendly application, CashVisionV2. The application serves a dual-system output, providing both text and voice outputs to ensure accessibility. The primary objective of this initiative is to enhance the financial independence of visually impaired individuals by simplifying their day-to-day transactions, particularly in the recognition of various Malaysian banknote denominations.

Actions (login required)