Research map / Image generation / Fast diffusion sampling
PolyDiffuse: Polygonal Shape Reconstruction via Guided Set Diffusion Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add guided set diffusion refinement to RoomFormer* proposals | RoomFormer* · Structured3D test | Angle F1: 83.9 -> 89.1 | Helped |
| Add guided set diffusion refinement to MapTR+ proposals | MapTR+ · nuScenes | Ordered-angle mAP: 43.8 -> 49.0 | Helped |
| Add guided set diffusion refinement to MapTR+ proposals | MapTR+ · nuScenes | Chamfer mAP: 59.3 -> 59.7 | No clear effect |
| Use rough annotations rather than MapTR proposals | PolyDiffuse · nuScenes | Ordered-angle mAP: 49.0 -> 45.3 | Hurt |
| Remove guidance networks and proposal initialization, using standard DM (ablation) | PolyDiffuse · Structured3D test | Angle F1: 89.1 -> 81.1 | Hurt |
| Reduce standard DM sampling from 10 to 2 steps (ablation) | Standard DM · Structured3D test | Angle F1: 81.1 -> 5.4 | Hurt |
| Increase PolyDiffuse training schedule from 1× to 2.5× (ablation) | PolyDiffuse · Structured3D test | Room F1: 98.0 -> 98.4 | Helped |
| Add intra-element attention to the denoising decoder (ablation) | PolyDiffuse · Structured3D test | Angle F1: 88.6 -> 89.1 | Helped |
| Use rough annotations instead of RoomFormer* proposals | PolyDiffuse · Structured3D test | Angle F1: 83.9 -> 89.4 | Helped |
Closest papers by meaning
- Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates (NeurIPS 2025 · similarity 0.88)
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.87)
- GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors (CVPR 2024 · similarity 0.87)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.87)
- Sin3DM: Learning a Diffusion Model from a Single 3D Textured Shape (ICLR 2024 · similarity 0.87)
- Motion2VecSets: 4D Latent Vector Set Diffusion for Non-rigid Shape Reconstruction and Tracking (CVPR 2024 · similarity 0.87)
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.