Development of Embedded System for Recognizing Kuzushiji by Deep Learning

Abstract

Effective utilization of “Pre-modern Japanese book database” constructed by the project supervised by Center for Collaborative Research on Pre-Modern Texts, National Institute of Japanese Literature, will push forward the development of the inter-field study. It may become an obstacle for the researchers with a little knowledge of classical literature, however, because historical Japanese texts have been written by Kuzushiji (Hentaigana and cursive kanji). In this article, we report an attempt of recognizing Kuzushiji by deep learning. Using Raspberry Pi which is the low-cost small single board computer, we developed the embedded system with enough recognizing rate and speed. In addition, manifold images of Kuzushiji written in a text image of the spread can be extracted. Our system will be effective for school children or elderly people not possessing mobile devices who want to come in touch with Kuzushiji.