Research map / 3D vision / 3D generation
Estimating Generic 3D Room Structures from 2D Annotations
Denys Rozumnyi, Stefan Popov, Kevis-kokitsi Maninis, Matthias Nießner, Vittorio Ferrari
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 automatic quality control filter (ablation) | Ours (R=100) · RE10k | Reprojection IoU: 0.83 -> 0.89 | Improved |
| Add automatic quality control filter (ablation) | Ours (R=100) · ScanNet | depth error (m): 0.30 -> 0.22 | Improved |
| Increase number of runs R from 1 to 100 (no quality control) (ablation) | Ours · RE10k | Reprojection IoU: 0.72 -> 0.83 | Improved |
| Increase number of runs R from 1 to 100 (no quality control) (ablation) | Ours · ScanNet | depth error (m): 0.36 -> 0.30 | Improved |
| Add perpendicularity loss L_P (ablation) | Ours (R=10) · RE10k | Reprojection IoU: 0.80 -> 0.90 | Improved |
| Add perpendicularity loss L_P (ablation) | Ours (R=10) · ScanNet | depth error (m): 0.34 -> 0.33 | No clear effect |
| Add edge loss L_E (ablation) | Ours (R=10) · RE10k | Reprojection IoU: 0.72 -> 0.90 | Improved |
| Add edge loss L_E (ablation) | Ours (R=10) · ScanNet | Reprojection IoU: 0.84 -> 0.84 | No clear effect |
| Add point tracking loss L_T (ablation) | Ours (R=10) · RE10k | Reprojection IoU: 0.34 -> 0.90 | Improved |
| Add point tracking loss L_T (ablation) | Ours (R=10) · ScanNet | depth error (m): Fail -> 0.33 | Improved |
| Train 2D layout estimator on the new dataset instead of LSUN | Lin et al. · pixel error | pixel error %: 6.25 -> 26.3 | Hurt |
Closest papers by meaning
- Structured3D: A Large Photo-realistic Dataset for Structured 3D Modeling (ECCV 2020 · similarity 0.89)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.89)
- GeoLayout: Geometry Driven Room Layout Estimation Based on Depth Maps of Planes (ECCV 2020 · similarity 0.88)
- M3DLayout: A Multi-Source Dataset of 3D Indoor Layouts and Structured Descriptions for 3D Generation (CVPR 2026 · similarity 0.88)
- PolyLayout: Multi-room Manhattan Layout Estimation (ECCV 2026 · similarity 0.88)
- No More Ambiguity in 360deg Room Layout via Bi-Layout Estimation (CVPR 2024 · 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.
