Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised Representation Learning for CAD
Benjamin T. Jones, Michael Hu, Milin Kodnongbua, Vladimir G. Kim, Adriana Schulz
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 |
|---|---|---|---|
| Pre-train face representations by reconstructing parametric surfaces and their clipping boundaries | B-Rep few-shot segmentation and classification · Fusion 360 Gallery Segmentation | accuracy | Improved |
| Use hierarchical message passing to add neighborhood context to self-supervised face embeddings (ablation) | Self-supervised B-Rep segmentation · Fusion 360 Segmentation and MFCAD | accuracy: 0.56 vs 0.55; 0.38 vs 0.35 (10 examples) | Improved |
| Use a nonlinear MLP decision boundary instead of a linear SVM (ablation) | Self-supervised B-Rep segmentation · Fusion 360 Segmentation and MFCAD | accuracy: 0.55 vs 0.50; 0.35 vs 0.15 (10 examples) | Improved |
| Replace the proposed self-supervision encoder with truncated SB-GCN (ablation) | B-Rep rasterization · Self-supervision reconstruction | XYZ Error: 0.0256 -> 0.035 | Hurt |
| Replace the proposed self-supervision encoder with truncated SB-GCN (ablation) | B-Rep rasterization · Self-supervision reconstruction | SDF Error: 0.0147 -> 0.0214 | Hurt |
| Add message passing before pooling (ablation) | B-Rep part classification · FabWave | accuracy: 0.72 vs 0.72; 0.97 vs 0.97; 1.00 vs 1.00 | No clear effect |
Closest papers by meaning
- FoV-Net: Rotation-Invariant CAD B-rep Learning via Field-of-View Ray Casting (CVPR 2026 · similarity 0.89)
- CADCrafter: Generating Computer-Aided Design Models from Unconstrained Images (CVPR 2025 · similarity 0.87)
- DTGBrepGen: A Novel B-rep Generative Model through Decoupling Topology and Geometry (CVPR 2025 · similarity 0.86)
- SpelsNet: Surface Primitive Elements Segmentation by B-Rep Graph Structure Supervision (NeurIPS 2024 · similarity 0.86)
- MiCADangelo: Fine-Grained Reconstruction of Constrained CAD Models from 3D Scans (NeurIPS 2025 · similarity 0.85)
- BrepGiff: Lightweight Generation of Complex B-rep with 3D GAT Diffusion (CVPR 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.
