Deep Learning-based 3D Reconstruction and Simulation Technology for Cheongsam and Hanbok

Chinese dress cheongsam is a treasure of Chinese culture, with the reality of online shopping and virtual cultural experience requirements, the virtual display and simulation technology of cheongsam and hanbok is also getting more and more attention. In this paper, we use GCN to deeply learn the deformation characteristics of cheongsam and hanbok, and successively realize the three-dimensional reconstruction of cheongsam and hanbok through pose estimation, feature line fitting and surface refinement. The spatio-temporal feature progressive fusion, multi-scale feature extraction and reconstruction modules are designed, and the fabric animation simulation method based on geometric images is proposed to enhance the display effect of cheongsam and hanfu. The reconstruction results of this paper’s method realize a more obvious improvement compared with all the reference models. The animation simulation error of cheongsam fabric is about 14% of PCA algorithm, and the time consumption is only about 2% of PBD algorithm, which verifies the feasibility of this paper’s work.