一种基于图卷积的网格生成方法
By designing a ChebConv-based residual network in the generator and using GAT in the discriminator, the information aggregation capability of the graph convolutional network is enhanced, solving the problem of lack of detail in the generator and achieving high-quality mesh generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUAIYIN INSTITUTE OF TECHNOLOGY
- Filing Date
- 2022-11-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing grid-based deep learning methods suffer from problems such as large memory consumption and lack of detail in the generated shapes when generating 3D shapes. Furthermore, the shallow structure of graph convolutional networks makes it difficult to effectively aggregate information from distant nodes.
A residual network based on ChebConv is used as the generator, and GAT is combined to enhance the information aggregation capability of the discriminator. A grid generation method is constructed through graph convolutional network to learn the different weights of surrounding nodes.
It improves the network's ability to aggregate information from more distant nodes, generates a grid model with more detailed features, and has significantly better training results than traditional methods.
Smart Images
Figure CN115761180B_ABST