Visual Communication of Ethnic Art Works on Social Networks Analyzed by Combining Deep Learning Models

Abstract

This paper proposes an analysis framework based on convolutional neural networks and generative adversarial networks to generate ethnic art works and verify their communication effects. First, a web crawler is written in Python to collect ethnic art works and their interactive data from social platforms, and a standardized dataset is constructed through data cleaning. Secondly, a convolutional neural network is used to extract the visual features of the works, such as patterns, colors, and structural information. The critical features related to the communication of ethnic art works are selected, and a dimensionality reduction is performed for input to the generative adversarial network. Then, a recurrent neural network is used to model the dynamic characteristics of the work’s communication, and a comprehensive prediction of the communication law is achieved through the LSTM model. By using the communication features extracted by RNN as the input conditions of the GAN generator, it guides the generation of artworks with high communication potential. Finally, ethnic art works are generated based on the generative adversarial network, and its effect on improving the communication depth and interaction rate in social networks is verified. The research results show that the number of likes for the GAN-generated content exceeds 1300. In the multi-dimensional comprehensive comparison, Multi-Angle Display has the highest comprehensive score, and the generated work content is the most attractive in the user feedback dimension, with the number of views reaching 98 points. The research results of this paper not only reveal the communication laws of ethnic art works, but also provide technical support and practical reference for improving their cultural influence.