Research map / 3D vision / Face and shape
Pseudo-Plane Regularized Signed Distance Field for Neural Indoor Scene Reconstruction
Jing Li, Jinpeng Yu, Ruoyu Wang, Shenghua Gao
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 sparse COLMAP depth supervision (ablation) | VolSDF · ScanNet (4 Manhattan scenes) | F-score: 0.397 -> 0.623 | Improved |
| add plane regularization (ablation) | VolSDF + Depth · ScanNet (4 Manhattan scenes) | F-score: 0.623 -> 0.657 | Improved |
| add plane segments fusion weighting (ablation) | VolSDF + Depth + Plane Reg · ScanNet (4 Manhattan scenes) | F-score: 0.657 -> 0.669 | Improved |
| add keypoint guided ray sampling (ablation) | VolSDF + Depth + Plane Reg + Weighting · ScanNet (4 Manhattan scenes) | F-score: 0.669 -> 0.683 | Improved |
| add keypoint guided sampling without plane regularization (ablation) | VolSDF + Depth · ScanNet (4 Manhattan scenes) | F-score: 0.623 -> 0.640 | Improved |
| use Gaussian rather than uniform sampling probability (ablation) | keypoint sampling · ScanNet (4 Manhattan scenes) | F-score: 0.617 -> 0.640 | Improved |
| use 8192 sample points instead of 1024 in plane estimation (ablation) | P2SDF · ScanNet (4 Manhattan scenes) | F-score: 0.633 -> 0.657 | Improved |
| increase sample points from 8192 to 16384 (ablation) | P2SDF · ScanNet (4 Manhattan scenes) | F-score: 0.657 -> 0.638 | Hurt |
| increase plane regularization loss weight from 0.2 to 1.0 (ablation) | P2SDF · ScanNet (4 Manhattan scenes) | F-score: 0.657 -> 0.592 | Hurt |
| segmentation loss weight 0.01 instead of 0.001 (ablation) | P2SDF · ScanNet (4 Manhattan scenes) | F-score: 0.649 -> 0.669 | Improved |
| add plane regularization (ablation) | Manhattan-SDF · ScanNet (4 Manhattan scenes) | F-score: 0.681 -> 0.688 | Improved |
| add plane weighting (ablation) | Manhattan-SDF + P · ScanNet (4 Manhattan scenes) | F-score: 0.688 -> 0.699 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- NeuS — NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction (NeurIPS 2021) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- NC-SDF: Enhancing Indoor Scene Reconstruction Using Neural SDFs with View-Dependent Normal Compensation (CVPR 2024 · similarity 0.90)
- ND-SDF: Learning Normal Deflection Fields for High-Fidelity Indoor Reconstruction (ICLR 2025 · similarity 0.88)
- NeRFPrior: Learning Neural Radiance Field as a Prior for Indoor Scene Reconstruction (CVPR 2025 · similarity 0.88)
- H2O-SDF: Two-phase Learning for 3D Indoor Reconstruction using Object Surface Fields (ICLR 2024 · similarity 0.86)
- Sparis: Neural Implicit Surface Reconstruction of Indoor Scenes from Sparse Views (AAAI 2025 · similarity 0.86)
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 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.
