Research On Chinese Calligraphy Style Recognition Method Based On Deep Learning

Calligraphy is the treasure of Chinese traditional culture.As one of the most popular cultural and artistic forms,calligraphy has been integrated into every aspect of modern life.The recognition of Chinese calligraphy style is of great significance in the fields of calligraphy imitation evaluation,learning appreciation and calligraphy font retrieval.Therefore,this thesis studies Chinese character detection and Chinese calligraphy style recognition algorithm based on deep learning technology,and constructs calligraphy style application by cascading the two modules.The main contents of this thesis are as follows:Firstly,in view of the lack of data sets,Chinese character detection data set and Chinese calligraphy style recognition data set were constructed by collecting and labeling ancient calligraphy works.In addition,Python and computer fonts were used to generate part of Chinese character detection data sets and standard calligraphy font data sets.Secondly,aiming at the problem of low detection accuracy of EAST algorithm model,the EAST benchmark model is improved from two aspects: feature extraction network and loss function.The feature extraction network in EAST model was replaced by Res Net and the Loss function was improved to Dice Loss.The accuracy rate of the improved model reached 88.4%,and the F1-Score reached 87.9%.Thirdly,aiming at the problem of low accuracy of Chinese calligraphy style recognition by existing methods,a Chinese calligraphy style recognition method based on Efficient Net was proposed.In this method,the original model structure is simplified and CBAM is introduced into the model to learn calligraphy features.The recognition accuracy of the improved model in the two data sets reached 98.68% and 76.64%.Finally,this thesis sets up the application of Chinese calligraphy style recognition.Based on the Vue framework and Python,the Chinese character detection algorithm and Chinese calligraphy style recognition algorithm are cascade together to realize the endto-end automatic recognition of Chinese calligraphy style.