02482nas a2200277   4500000000100000008004100001653002500042653002600067653002000093653001200113653002100125653002300146653002600169653003400195653002200229653002800251653001800279653005900297653002100356653002300377100001500400700001900415245015600434856015900590520145500749                                       d10aAdversarial networks10aChinese paper cutting10aComputer Vision10aCutting10aCutting patterns10aDigital protection10aHistoric preservation10aIntangible cultural heritages10aMean square error10aOptimization algorithms10aPaper cutting10aProgressive conditional generative adversarial network10aRealistic sample10aVision transformer1 aHanwen Fan1 aXiaomeng Zhang00aDigital Protection and Inheritance of Intangible Cultural Heritage using Progressive Conditional Generative Adversarial Network with Vision Transformer  uhttps://www.scopus.com/inward/record.uri?eid=2-s2.0-105011827817&doi=10.1109%2FICDSIS65355.2025.11071120&partnerID=40&md5=6e354837a4675343fa6dda6a1e25cdde3 aSince 21st Century, Chinese paper cutting has been recognized as an important element of Intangible Cultural Heritage (ICH) of China, which served as a symbol of traditional aesthetics, cultural identity and social values. Due to their rich cultural importance, the digital protection of paper cutting artwork is crucial. Moreover, existing approaches such as Progressive Conditional Generative Adversarial Network with Archimedes Optimization Algorithm (PCGAN-AOA) struggle from generalizability, due to lack of ability to capture long range dependencies in paper cutting patterns. Hence, A hybrid novel approach namely, PCGAN with Vision Transformer (PCGAN-ViT) is proposed for digital protection of ICH and inheritance of Chinese paper cutting. Initially, the input data consists of complete paper cutting pattern images which are gathered from multiple sources such as local cultural heritage institutions, museums, and verified public repositories. Then these images are preprocessed by performing resizing through bilinear interpolation and normalization of pixels through min max scaling. Lastly, these preprocessed images are fed to the proposed PCGAN-ViT model, where PCGAN performs the generation of realistic samples and ViT is used to verify the stylistic correctness by classifying the generated patterns. Therefore, the proposed PCGAN-ViT achieved low Root Mean Squared Error (RMSE) of 0.097 which outperformed existing PCGAN-AOA model.