A Study on Packaging Design Style Transfer Based on CBAM and GAN

The packaging design industry, as a critical field in product marketing and brand image development, has continuously evolved in response to market demands and technological advancements. At present, artificial intelligencedriven style transfer methods are rapidly integrating into the packaging design sector. By intelligently deconstructing and reconstructing image content and styles, these methods provide a more efficient, diverse, and personalized approach to innovation in packaging design. Current style transfer methods have shortcomings in detail preservation and global style consistency. Therefore, this study proposes a style transfer model based on convolutional block attention and cycleconsistent generative adversarial networks (GANs). After 200 iterations, the proposed model maintains a high accuracy of 84.5%, with a loss rate and recall of 5.8% and 86.7%, respectively. Following tests on 500 image samples, the model achieves a Fréchet Inception Distance of 12.3. Experimental results indicate that the proposed model outperforms the comparison models in preserving brand logos and complex graphic textures, with notable improvements in both local detail retention and overall style consistency. The model provides a highly practical solution for packaging design and contributes to advancing intelligent design technologies in commercial applications.