The intangible cultural heritage art is challenged by digital transformation issues such as fragmented data of pattern, inability to measure the stylistic features and ineffectiveness in interaction. To overcome these problems, this paper develops a pattern generation and interaction design system to be based on intelligent algorithm. Pattern morphological features are extracted with the help of a Convolutional Neural Network (CNN) and multi-style fusion generation is performed with the help of Generative Adversarial Network (GAN). The control of feature hierarchical weight is added in the transfer of styles in order to retain details and provide the maximization of the congruency of styles. The interaction design element applies Deep Reinforcement Learning (DRL) to develop a user feedback model to improve the individualized creative experience by modifying the parameters dynamically. The Python and TensorFlow are used to implement the system, which is trained on 2000 pattern images. The best performance is in generation time which is clustered at 0.8 to 0.95 seconds per image with an 8.5 to 9.1 subjective interaction convenience score. The maximum SSIM score of 0.956 is in the style reproduction accuracy. The findings of the research indicate that the algorithm-based digital approach is capable of producing the desired high quality of generation and the interactive presentation of the art patterns of the intangible cultural heritage, which provides confirmation of the possibility of its application in the sphere of cultural inheritance and redesign.
Jing Yu (Thu,) studied this question.