Chinese calligraphy recognition system based on convolutional neural network

This paper presents a Chinese calligraphy recognition system which can ex- tract text from an image of Chinese calligraphy and recognize its style of calligraphy fonts. A multi-label convolutional neural network (CNN) recognition model is created and trained to recognize both textual content and font of single Chinese character at the same time. A large number of calligraphy images of single Chinese character are collected and prepossessed to form a dataset for training the model. Several images of calligraphy works of ancient Chinese calligraphers are used to evaluate the performance of the pro- posed system, and the experimental results showed the capability of the proposed system to recognize Chinese calligraphy.