Design of automatic recognition and intelligent generation system for China traditional patterns based on convolutional neural network

In this paper, an automatic identification and intelligent generation system of China traditional patterns based on convolutional neural network (CNN) is proposed, which realizes the efficient identification and innovative design of traditional patterns through modern AI technology and promotes the digital protection and inheritance of cultural heritage. The overall architecture of the system includes data preprocessing module, pattern automatic recognition module, pattern intelligent generation module and user interaction interface. In the aspect of automatic pattern recognition, a dual-path feature fusion model based on ResNet-50 backbone network is designed. By combining depth separable convolution and hole convolution with attention mechanism, multi-scale structural features and local texture features are efficiently extracted, which significantly improves the recognition accuracy. The experimental results show that this model achieves an accuracy of 94.7% and a F1 value of 0.932 on the data set of the Palace Museum, which is superior to the classical models such as ResNet-50 and VGG16. In the aspect of pattern intelligent generation, an improved StyleGAN2 architecture (TGAN) is constructed. Through the conditional generation module and multi-scale generator, high-quality pattern images with traditional charm are generated by combining the traditional pattern template library and grammar rule library. The objective evaluation index shows that the FID value of TGAN model IS 18.2, the IS value is 3.89, and the SSIM value is 0.83, all of which are better than the comparison model, and it is highly recognized by experts in terms of cultural compliance. This study provides a new technical path for the digital protection and innovative design of traditional patterns, which has important cultural and application value.