Thai Currency Classification Application for Visually Impaired Persons

Abstract

This research presents an application designed to assist visually impaired individuals in identifying Thai banknotes and coins. The application utilizes image processing and machine learning techniques, specifically a Convolutional Neural Network with ResNet101 architecture, to accurately classify 11 types of Thai currency. It is designed for offline use on smartphones, providing real-time audio and text output to enhance accessibility and understanding for users with visual impairments. The dataset includes 2,593 images of Thai banknotes and coins, split into 80% for training and 20% for testing. The application employs the trained model to conduct real-world tests using a smartphone camera, testing with actual banknotes and coins, achieving an average accuracy of 92.73%.