Research on Applying Dense Convolutional Neural Network in Chinese Character Font Recognition

The emergence of computer and artificial intelligence technology has subverted the traditional work model of pattern recognition and classification. Chinese character font recognition is one of the essential steps in developing and disseminating digitalized traditional calligraphy. In the traditional machine learning environment, dealing with the automatic recognition of a large number of fonts is often challenging. At the same time, font recognition under manual intervention also has substantial limitations. With the development of deep learning, the construction of font recognition algorithms using convolutional neural networks has become a mainstream research direction. This paper proposes a dense convolutional neural network, especially using fonts such as regular script, cursive script, and seal character as training targets to construct a mixed training set. By optimizing the model in three aspects: pooling rules, training strategies, and model pruning, the recognition performance of the model is further improved. It is hoped that the research in this paper will provide some help for the recognition of Chinese characters and the dissemination of digitalized traditional calligraphy.