Abstract

Chinese calligraphy is globally esteemed for its therapeutic and mindfulness benefits, encompasses foundational scripts such as regular (Kai Shu), running (Xing Shu), official (Li Shu), and cursive (Cao Shu). Beginners typically master the structured regular script before advancing to the complexities of cursive forms. Each aesthetic style is defined by unique historical contributions, requiring learners to discern subtle stylistic nuances. This study introduces an innovative convolutional neural network (CNN) architecture, pioneering the application of deep learning in the classification of these calligraphic traditions. Focusing on the four primary scripts from the Tang Dynasty (690–907 A.D.), a period marking the zenith of regular script refinement, a comprehensive dataset of 8,282 samples was compiled for model training. Our approach distinguishes individual artistic styles with superior performance over existing networks, achieving 89.5–96.2% accuracy. This research highlights the efficacy of CNNs in categorizing both font and artistic styles, advancing cultural and technical calligraphy studies.