Toward Script Translation in Calligraphic Works Using Deep Learning

–Character recognition of seal scripts–

Abstract

Calligraphic scripts are classified broadly into five types: seal, clerical, cursive, running, and standard. In this study, we propose a convolutional neural network–based method for character recognition of seal scripts. Utilizing the feature that kanji is composed of parts such as the radicals, we can recognize unreadable characters used in calligraphic works through image recognition of each part of the characters. Our method provides a general purpose processing environment of character recognition with small numbers of data and classes.