Research map / Architectures and efficiency / RNNs and normalization
SketchGen: Generating Constrained CAD Sketches
Wamiq Reyaz Para, Shariq Farooq Bhat, Paul Guerrero, Tom Kelly, Niloy Mitra, Leonidas Guibas, Peter Wonka
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 SketchGen instead of SG-sketch | Sketch generation · SketchGraphs test set | NLL/sketch: 158.90 -> 88.22 | Improved |
| Use SketchGen instead of SG-constraint | Auto-constraining · SketchGraphs test set | NLL/sequence: 19.29 -> 8.28 | Improved |
| Use nucleus sampling p=0.9 instead of p=1.0 | SketchGen full sketch generation · SketchGraphs generated sequences | E_syntax: 19.8 -> 18.3 | Improved |
| Use nucleus sampling p=0.9 instead of p=1.0 | SketchGen full sketch generation · SketchGraphs generated sequences | E_stat: 0.0192 -> 0.0627 | Hurt |
| Use nucleus sampling p=0.9 instead of p=1.0 | SketchGen auto-constraining · SketchGraphs generated sequences | E_syntax: 1.56 -> 1.29 | Improved |
| Use nucleus sampling p=0.9 instead of p=1.0 | SketchGen auto-constraining · SketchGraphs generated sequences | E^C_stat: 0.0415 -> 0.0442 | Hurt |
| Add Q3 syntax input (ablation) | SketchGen primitive generator · SketchGraphs primitive model test set | NLL/sequence: 79.96 -> 79.94 | No clear effect |
| Add Q3 syntax input (ablation) | SketchGen primitive generator · SketchGraphs primitive model test set | NLL/primitive: 8.605 -> 8.600 | No clear effect |
| Add Q3 and Q4 syntax inputs (ablation) | SketchGen primitive generator · SketchGraphs primitive model test set | NLL/sequence: 80.35 -> 79.94 | Improved |
| Add Q4 syntax input (ablation) | SketchGen constraint generator · SketchGraphs auto-constraining test set | NLL/sequence: 8.47 -> 8.28 | Improved |
| Group embeddings by parameter type instead of mixing token types (ablation) | SketchGen constraint generator · SketchGraphs auto-constraining test set | NLL/sequence: 8.45 -> 8.28 | Improved |
| Use context-dependent learned constraint generation instead of distance-threshold snapping | Auto-constraining · SketchGraphs test sketches | coincident-constraint false positives | Improved |
Closest papers by meaning
- Revisiting CAD Model Generation by Learning Raster Sketch (AAAI 2025 · similarity 0.89)
- SketchDNN: Joint Continuous-Discrete Diffusion for CAD Sketch Generation (ICML 2025 · similarity 0.87)
- Text2CAD: Generating Sequential CAD Designs from Beginner-to-Expert Level Text Prompts (NeurIPS 2024 · similarity 0.87)
- Generating CAD Code with Vision-Language Models for 3D Designs (ICLR 2025 · similarity 0.87)
- MiCADangelo: Fine-Grained Reconstruction of Constrained CAD Models from 3D Scans (NeurIPS 2025 · similarity 0.87)
- FlexCAD: Unified and Versatile Controllable CAD Generation with Fine-tuned Large Language Models (ICLR 2025 · similarity 0.86)
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.
