Kanji recognition with AI

Abstract

This thesis explores the development of a mobile application for recognizing Kanji characters using artificial intelligence (AI) techniques. Beginning with an introduction to the Japanese language and an overview of Kanji, the thesis continues with the fundamentals of AI, including its types and applications in various domains. It discusses machine learning principles, focusing on supervised and unsupervised learning, along with error metrics such as overfitting and underfitting. The thesis then delves into deep learning concepts, including neural networks, activation functions, and training methodologies. Special attention is given to convolutional neural networks (CNNs) due to their efficacy in image recognition tasks. Moreover, it examines standard tools used in AI applications, particularly Python programming language and associated libraries. The thesis culminates in designing and implementing a Japanese kanji Recognition mobile app, detailing its frontend and backend components.