Research map / 3D vision / Gaussian splatting
Learning Scene-Level Signed Directional Distance Function With Ellipsoidal Priors and Neural Residuals
Zhirui Dai, Hojoon Shin, Yulun Tian, Ki Myung Brian Lee, Nikolay Atanasov
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 negative SDDF sample augmentation (ablation) | SDDF · Replica Hotel (LiDAR) | MAE (cm): 2.639 -> 0.997 | Improved |
| add negative SDDF sample augmentation (ablation) | SDDF · Gibson Allensville (LiDAR) | MAE (cm): 5.379 -> 1.350 | Improved |
| use polynomial embedding instead of Fourier embedding (ablation) | SDDF · Gibson Allensville (LiDAR) | MAE (cm): 1.842 -> 1.350 | Improved |
| use hybrid initialization instead of K-means++ only (ablation) | SDDF · Gibson Allensville (LiDAR) | MAE (cm): 1.405 -> 1.350 | Improved |
| double the number of ellipsoids (ablation) | SDDF · Gibson Allensville (LiDAR) | MAE (cm): 1.350 -> 1.414 | Hurt |
| halve the number of ellipsoids (ablation) | SDDF · Gibson Allensville (LiDAR) | MAE (cm): 1.350 -> 1.477 | Hurt |
| vary ellipsoid count (half or twice) (ablation) | SDDF · Office 2 (LiDAR) | MAE (cm): 1.236 -> 1.218 (half), 1.291 (twice) | Mixed |
| train with depth loss | Nerfacto · Allensville (RGB-D) | MAE (cm): 83.433 -> 85.655 | No clear effect |
| remove RGB input | RaDe-GS · Allensville | MAE (cm): 1.737 -> 25.002 | Hurt |
| SDDF vs SDF-Instant-NGP with sphere tracing, inference time | SDDF · Replica Hotel | inference time (ms/frame): 103 -> 69 | Improved |
| SDDF vs RaDe-GS training GPU memory | SDDF · Replica Hotel | training GRAM (G): 18.7 -> 3.3 | Improved |
| use only depth from RGB-D vs LiDAR input | SDDF · Gibson Allensville | MAE (cm): 1.350 -> 1.490 | Hurt |
Closest papers by meaning
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 2024 · similarity 0.91)
- VRP-UDF: Toward Unbiased Learning of Unsigned Distance Functions From Multi-View Images With Volume Rendering Priors (TPAMI 2026 · similarity 0.91)
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.90)
- GSDF: 3DGS Meets SDF for Improved Neural Rendering and Reconstruction (NeurIPS 2024 · similarity 0.90)
- NC-SDF: Enhancing Indoor Scene Reconstruction Using Neural SDFs with View-Dependent Normal Compensation (CVPR 2024 · similarity 0.89)
- NeRFPrior: Learning Neural Radiance Field as a Prior for Indoor Scene Reconstruction (CVPR 2025 · similarity 0.89)
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.
