Research map / Graph learning / Graph self-supervised learning
BrepVGAE: Variational Graph Autoencoder with Unified Latent Representation for B-rep
Hao Guo, Liyuan Deng, Yongkang Dai, Ruohan Wang, Jiahao Li, Yunpeng Bai, Yilei Shi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use SparseGNN instead of GAT (ablation) | BrepVGAE encoder · DeepCAD generation (60K ablation) | COV: 60.46 -> 78.75 | Improved |
| use SparseGNN instead of GCN (ablation) | BrepVGAE encoder · DeepCAD generation (60K ablation) | COV: 68.78 -> 78.75 | Improved |
| use SparseGNN instead of GCN (ablation) | BrepVGAE encoder · DeepCAD generation (60K ablation) | MMD: 1.18 -> 1.18 | No clear effect |
| use multi-slot learnable queries with KV expansion vs single query (ablation) | set decoder · DeepCAD generation (60K ablation) | COV: 65.30 -> 78.75 | Improved |
| add key-value expansion (ablation) | set decoder · DeepCAD generation (60K ablation) | COV: 70.25 -> 78.75 | Improved |
| make queries learnable vs fixed (ablation) | set decoder · DeepCAD generation (60K ablation) | COV: 63.53 -> 78.75 | Improved |
| two-stage training vs end-to-end (ablation) | BrepVGAE · DeepCAD generation (60K ablation) | COV: 60.20 -> 78.75 | Improved |
| two-stage training vs frozen geometry throughout (ablation) | BrepVGAE · DeepCAD generation (60K ablation) | COV: 76.35 -> 78.75 | Improved |
| add CD loss to edge autoencoder (ablation) | Edge VAE · Edge point cloud AE | CD: 0.987 -> 0.658 | Improved |
| add CD loss to face autoencoder (ablation) | Face VAE · Surface point cloud AE | CD: 1.126 -> 0.619 | Improved |
| Hungarian vs nearest-neighbor matching | Ours · DeepCAD generation | COV: 73.40 -> 79.82 | Improved |
| Hungarian vs Sinkhorn matching | Ours · DeepCAD generation | COV: 78.71 -> 79.82 | Improved |
Closest papers by meaning
- HiFi-BRep: High-Fidelity Latent Representation for Robust B-Rep Generation (CVPR 2026 · similarity 0.89)
- BrepGiff: Lightweight Generation of Complex B-rep with 3D GAT Diffusion (CVPR 2025 · similarity 0.88)
- DTGBrepGen: A Novel B-rep Generative Model through Decoupling Topology and Geometry (CVPR 2025 · similarity 0.88)
- Graph Geometry-Preserving Autoencoders (ICML 2024 · similarity 0.86)
- GraphCroc: Cross-Correlation Autoencoder for Graph Structural Reconstruction (NeurIPS 2024 · similarity 0.86)
- Teacher-guided Edge Discriminator for Personalized Graph Masked Autoencoder (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
