Agricultural and sideline products in Jilin Province lack distinctive image identity, AI image generation technology provides a new way to create distinctive comic IP characters, which helps to enhance brand value and promote rural revitalization. In this study, we constructed an intelligent image generation model of Multi-shape Generative Adversarial Network (MW-GAN), which contains two parts: style network and geometric shape network, and ensures that the features of agricultural and sideline products are retained through techniques such as identity preserving loss design. The experiment uses a homemade agricultural and by-products caricature dataset (AVPDS), which contains a total of 2,165,496 images of 3,214 agricultural and by-products from 10 angles in Jilin Province. The evaluation results show that the IS index value of the model reaches 94.91, which improves from 7.46% to 41.03% compared with the comparison method; the FID index value is 6.63, which reduces from 6.88% to 26.59% compared with the comparison method. In extreme postures (75° and 90°), the recognition rates reached 85.45% and 77.83%, respectively. The user survey showed that the agricultural and sideline products cartoon IP image value score was 4.319, the brand association score was 4.162, and the brand identity score was 4.285, and the regression analysis confirmed that all three factors had a significant positive impact on rural revitalization. The results of the study show that the design of agricultural and sideline products cartoon IP image based on AI technology can effectively disseminate brand culture, enhance product recognition, and promote the development of rural revitalization.
