The Optimization of Illustration Design in Cultural and Creative Products for Liaoning Region Under Intelligent Generative Adversarial Network

This study explores the application of intelligent Generative Adversarial Networks (GANs) in illustration design and cultural and creative product design in Liaoning. It aims to enhance design efficiency and quality while promoting the inheritance and innovation development of local cultural heritage. The study employs CycleGAN combined with self-attention mechanisms to conduct style transfer experiments on cultural illustrations featuring Liaoning’s regional characteristics. It compares the performance of six models, including traditional GAN, CycleGAN, and CycleGAN + Self-attention, in terms of image quality and training efficiency. The results show that the CycleGAN + Self-attention model outperforms others in image quality indicators (PSNR and SSIM). On the Liaoning illustration dataset, this model achieves a PSNR of 30.5 dB and an SSIM of 0.96, which are 10 dB and 0.21 higher than traditional GANs. Tests across diverse datasets also demonstrate its robust generalization and adaptability. The proposed CycleGAN + Self-attention model performs well in illustration and cultural product design, generating high-quality illustrations with local cultural features. Beyond enhancing design efficiency, this model offers new technical means for cultural inheritance. Overall, this study furnishes novel ideas and methods for integrating intelligent GANs into art design practices. It helps promote Liaoning’s cultural heritage and innovation, and supports the development of the cultural and creative industry.