Research map / 3D vision / NeRF and radiance fields
Surface Normal Clustering for Implicit Representation of Manhattan Scenes
Nikola Popovic, Danda Pani Paudel, Luc Van Gool
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 Manhattan self-supervision (L_ort + L_ctr) to InstantNGP (ablation) | InstantNGP · Hypersim-A | PSNR: 25.86 -> 27.20 | Improved |
| add L_ctr cluster-tightening loss (ablation) | InstantNGP · Hypersim-A | normal error (deg): 57.12 -> 36.09 | Improved |
| add L_ort orthogonality loss (ablation) | InstantNGP · Hypersim-A | rotation error (deg): 10.63 -> 5.39 | Improved |
| provide the Manhattan frame (MF known) instead of learning it (ablation) | Ours · Hypersim-A | PSNR: 27.20 -> 27.21 | No clear effect |
| delay the Manhattan losses until step 500 with linear weight ramp (ablation) | Ours · Hypersim-A | PSNR: 27.06 -> 27.20 | Improved |
| replace ManhattanDF (needs semantics and floor axis) with label-free self-supervision | InstantNGP backbone · ScanNet | PSNR: 18.68 -> 20.79 | Improved |
| replace ManhattanDF with label-free self-supervision on easy scenes | InstantNGP backbone · Replica | PSNR: 35.24 -> 35.13 | No clear effect |
| apply method with sparse training views | InstantNGP · Hypersim-A, 6 views | PSNR: 15.75 -> 16.67 | Improved |
| increase triplet triangle size | Ours · Hypersim-A | PSNR / MF estimation | Hurt |
| sample random triplet triangles instead of random rays for the baseline (ablation) | InstantNGP · Hypersim-A | PSNR: 25.86 -> 26.16 | No clear effect |
Closest papers by meaning
- Mask-Based Modeling for Neural Radiance Fields (ICLR 2024 · similarity 0.88)
- Semantically-aware Neural Radiance Fields for Visual Scene Understanding: A Comprehensive Review (IJCV 2026 · similarity 0.87)
- UNISURF: Unifying Neural Implicit Surfaces and Radiance Fields for Multi-View Reconstruction (ICCV 2021 · similarity 0.87)
- NC-SDF: Enhancing Indoor Scene Reconstruction Using Neural SDFs with View-Dependent Normal Compensation (CVPR 2024 · similarity 0.87)
- NeRFPrior: Learning Neural Radiance Field as a Prior for Indoor Scene Reconstruction (CVPR 2025 · similarity 0.87)
- Latent Radiance Fields with 3D-aware 2D Representations (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.
