Optimization method of AIGC art design creation system integrating deep learning and image processing

In this paper, an optimization method of AIGC (Artificial Intelligence Generated Content) art design creation system combining deep learning and image processing is proposed. Through algorithm optimization, hyperparameter adjustment, training strategy improvement and the introduction of user feedback mechanism, the performance and user experience of the system are improved. In the optimization method, this paper realizes the creative enhancement of multi-scale information integration and generation algorithm by loading pre-training model, configuring dynamic batch size and adaptive learning rate, and combining feature fusion layer and creative excitation layer. Hyperparameter optimization adopts learning rate attenuation strategy and dynamic batch size selection, which effectively balances training efficiency and model performance. The improved early stop method and hierarchical transfer learning strategy further avoid over-fitting and reduce training time. In the experimental part, this paper uses WikiArt-10k and COCO-Art data sets for training, and uses FID, PSNR, SSIM and other objective indicators and user satisfaction surveys to evaluate. The results show that the optimized model has significantly improved the image quality, training efficiency and user satisfaction, with FID decreased by 33.3%, PSNR and SSIM increased by 15.8% and 9.0% respectively, training time decreased by 23%, and user satisfaction with artistic creativity and style diversity increased by 21.3% and 24.4% respectively. The ablation experiment further verified the effectiveness of each optimization module. The research in this paper provides a useful reference for the development of AIGC art design creation system, which can promote the wide application of AIGC technology in the field of art design, provide artists and creators with more efficient and diversified creative tools, and promote artistic innovation and inheritance.