Research on the Application of Deep Learning Hierarchical Perception Technology in 3D Garment Design

This paper designs an innovative hierarchical awareness architecture, which combines the powerful graph structure learning ability of GCN and the flexible information screening advantages of the attention mechanism. The designed hierarchical awareness architecture is divided into two stages: local feature extraction and global information integration. In the local feature extraction stage, GCN performs a convolution operation on each vertex of the 3D mesh to learn its neighborhood information. Subsequently, the attention mechanism evaluates the importance of each vertex, assigning different weights to emphasize key design elements. In the global information integration stage, the weighted features of all vertices are aggregated through the multi-head attention mechanism to form a comprehensive design representation, which not only contains local details, but also considers the consistency and coordination of the overall design. The combination of GCN and the attention mechanism enables the model to control the clothing mesh finely, and realize high-fidelity 3D clothing design.