Research map / Label-efficient and robust learning / Recognition and long-tail
Discovering Design Concepts for CAD Sketches
Yuezhi Yang, Hao Pan
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 concept-based generation rather than autoregressive generation | CAD sketch image-conditioned generation · Image-conditioned generation | Primitive F-score: 0.575 -> 0.711 | Improved |
| use concept-based generation rather than autoregressive generation | CAD sketch image-conditioned generation · Image-conditioned generation | Constraint F-score: 0.301 -> 0.368 | Improved |
| add binary reference costs to reconstruction loss (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Constraint F-score: 0.290 -> 0.766 | Improved |
| add sharp reference loss (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Constraint F-score: 0.104 -> 0.766 | Improved |
| add modularity enhancement bias loss (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Modular(%): 13.3 -> 50.8 | Improved |
| increase concept query count from 5 to 12 (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Constraint F-score: 0.766 -> 0.918 | Improved |
| increase concept query count from 5 to 12 (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Modular(%): 50.8 -> 14.9 | Hurt |
| increase argument count from 2 to 4 (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Modular(%): 50.8 -> 7.1 | Hurt |
| decrease argument count from 2 to 1 (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Constraint F-score: 0.766 -> 0.666 | Hurt |
| increase concept library size from 50 to 2000 (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Constraint F-score: 0.581 -> 0.779 | Improved |
| increase concept library size from 50 to 2000 (ablation) | Sketch concept induction · SketchGraphs auto-encoding | Modular(%): 48.1 -> 50.6 | No clear effect |
Closest papers by meaning
- SketchDNN: Joint Continuous-Discrete Diffusion for CAD Sketch Generation (ICML 2025 · similarity 0.85)
- CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention (CVPR 2024 · similarity 0.85)
- Text2CAD: Generating Sequential CAD Designs from Beginner-to-Expert Level Text Prompts (NeurIPS 2024 · similarity 0.85)
- Revisiting CAD Model Generation by Learning Raster Sketch (AAAI 2025 · similarity 0.84)
- MiCADangelo: Fine-Grained Reconstruction of Constrained CAD Models from 3D Scans (NeurIPS 2025 · similarity 0.84)
- From 2D CAD Drawings to 3D Parametric Models: A Vision-Language Approach (AAAI 2025 · similarity 0.84)
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.
