Transformer-based Generation of Confrontation Network in Digital Art Applications
Huiping Meng, Feng Gao, Dong Li, Yue Liu, Jianhui Xu, Mengjiao Wang, Jian Xiang Yang · 2024
With the continuous progress and development of society, the application of generative adversarial network (GAN) in digital art is increasing day by day. As a deep learning model, Transformer can bring better help to the development of GAN. This article studies the application of Transformer-based GAN in digital art, and the focus of the research is to introduce its achievements in style migration and image restoration. First, it uses Transformer to build a GAN model, and then applies the built model to style migration and image repair to achieve high-quality and diverse style migration effects, as well as effective image repair. Experiments have proved that the value of IS (Inception Score) studied in this article is higher than that of other models, at 3.260.561, while the value of FID (Fréchet Inception Distance) is lower than that of other models, at 66.81. By combining the efficient processing capability of Transformer with the generation capability of GAN, higher quality and more personalized digital art creation can be achieved.