Research on classification of woodblock New Year pictures based on ResNet improved model

In order to solve the problem that traditional manual classification methods are cumbersome and time-consuming, deep learning is used to classify images of intangible cultural heritage woodblock New Year paintings. Reduce time-consuming problems, effectively improve the accuracy of classification, and maximize the improvement and optimization of the digital woodblock New Year painting inheritance system. In order to more accurately classify various woodblock New Year pictures, the original ResNet was improved to more comprehensively analyze the detailed features of the woodblock New Year pictures and improve the classification effect of the model. The experiment selected 12 types of woodblock New Year pictures, with a total of 2160 original woodblock New Year pictures. By comparing with multiple other methods, experimental results show that the accuracy of the improved model based on ResNet reaches $\mathbf{9 1 . 0 5 \%}$. It has better effect on classifying various woodblock New Year pictures.