Exploring the Application of Artificial Intelligence in Design Courses in Colleges and Universities: an Example from Graphic Design

Artificial intelligence graphic generation can realize the generation of graphic works with human aesthetic beauty and design principles by simulating the creation process, which helps all aspects of design professional teaching. This paper constructs a graphic design teaching model based on generative adversarial networks. The residual attention multi-channel generative adversarial network is proposed. First, residual connections are utilized so that the transmission of model information becomes more efficient and model training becomes more stable. Second, the self-attention mechanism is introduced to strengthen the long-distance dependence of the generated images and improve the model’s ability to model shapes. The effectiveness and generation results of the model are analyzed using the design text dataset and the graphic dataset. The model is applied to assist in the teaching of design majors in a university to analyze its effect on improving teaching efficiency. It is found that the performance of this paper’s model in the evaluation of node degree distribution, clustering coefficient distribution, and average number of orbits distribution is still improved after more than three layers, and the enhancement of residual connectivity is confirmed. Compared with the baseline model, this paper’s model generates results with MMD values lower than 0.20 on all datasets.In the teaching experiment, the average score of the experimental group is 7.28 points higher than that of the control group, and there is a significant difference in 9 out of the 12 items examined. This paper contributes new ideas and methods to improve the teaching efficiency of design majors and achieve personalized design practice.