Research map / 3D vision / Face and shape
Neural Procedural Reconstruction for Residential Buildings
Huayi Zeng, Jiaye Wu, Yasutaka Furukawa
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 |
|---|---|---|---|
| add dormer augmentation to training data (ablation) | NPR dormer detection · Dormer detection (3 px) | precision and recall: (8% and 17%) to (77% and 84%) | Improved |
| add synthetic data (x40) to real training data (ablation) | NPR foundation classification · Foundation classification | accuracy: 82.2 -> 89.2 | Improved |
| add synthetic data (x10) to real training data (ablation) | NPR foundation classification · Foundation classification | accuracy: 82.2 -> 85.3 | Improved |
| add more synthetic data (x50 vs x40) (ablation) | NPR foundation classification · Foundation classification | accuracy: 89.2 -> 88.7 | No clear effect |
| add synthetic data to real training data (ablation) | NPR II-foundation regression · II-foundation regression | IoU: 0.796 -> 0.815 | No clear effect |
| train on synthetic data only (x10 to x50) (ablation) | NPR foundation classification · Foundation classification | accuracy: 89.2 -> 42.0 | Hurt |
| train regression on synthetic data only instead of real plus synthetic (ablation) | NPR II-foundation regression · II-foundation regression | IoU: 0.815 -> 0.803 | Hurt |
| replace Poisson+QSlim reconstruction with NPR | residential LiDAR reconstruction · Residential LiDAR (250 houses) | mean absolute distance: 1.47 -> 0.81 | Improved |
| replace piecewise planar reconstruction with NPR | residential LiDAR reconstruction · Residential LiDAR (250 houses) | mean absolute distance: 0.95 -> 0.81 | Improved |
| replace Kentucky bottom-up decomposition with NPR | residential LiDAR reconstruction · Residential LiDAR (250 houses) | mean absolute distance: 2.13 -> 0.81 | Improved |
| replace USC with NPR | residential LiDAR reconstruction · Residential LiDAR (250 houses) | number of triangles: 169.0 -> 45.5 | Improved |
| replace piecewise planar with NPR on U-foundation houses | residential LiDAR reconstruction · Residential LiDAR (U-foundation houses) | mean absolute distance: 0.80 -> 0.84 | Hurt |
Closest papers by meaning
- Visual-to-Executable Procedural Reconstruction of Buildings. (NeurIPS 2026 · similarity 0.89)
- FloorNet: A Unified Framework for Floorplan Reconstruction from 3D Scans (ECCV 2018 · similarity 0.87)
- ArcPro: Architectural Programs for Structured 3D Abstraction of Sparse Points (CVPR 2025 · similarity 0.86)
- Intelligent Home 3D: Automatic 3D-House Design From Linguistic Descriptions Only (CVPR 2020 · similarity 0.85)
- Connecting the Dots: Floorplan Reconstruction Using Two-Level Queries (CVPR 2023 · similarity 0.85)
- Building3D: A Urban-Scale Dataset and Benchmarks for Learning Roof Structures from Point Clouds (ICCV 2023 · 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.
