Research map / Point clouds / Shape and surface completion
RayletDF: Raylet Distance Fields for Generalizable 3D Surface Reconstruction from Point Clouds or Gaussians
Shenxing Wei, Jinxi Li, Yafei Yang, Siyuan Zhou, Bo Yang
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 |
|---|---|---|---|
| use multiple raylets (T=5) instead of one in training (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.174 -> 0.145 | Improved |
| add confidence score s_l to blender (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.170 -> 0.145 | Improved |
| use softmax-normalized confidence instead of sigmoid (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.164 -> 0.145 | Improved |
| use softmax-normalized confidence instead of alpha blending (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.156 -> 0.145 | Improved |
| include both xyz and relative position of neighbors (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.148 -> 0.145 | Improved |
| use K=5 neighbors instead of K=1 (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.151 -> 0.145 | Improved |
| increase neighbors from K=5 to K=20 (ablation) | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.145 -> 0.152 | Hurt |
| increase test-time raylets from T=1 to T=10 | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.177 -> 0.140 | Improved |
| increase test-time raylets from T=5 to T=20 | RayletDF · MultiScan (unseen) | ADE (m): 0.259 -> 0.247 | Improved |
| increase test-time raylets from T=10 to T=20 | RayletDF · ScanNet/ScanNet++ (3D Gaussians) | ADE (m): 0.140 -> 0.141 | No clear effect |
| replace RayDF with RayletDF, cross-dataset (train ScanNet, test MultiScan) | point cloud input · MultiScan | ADE (m): 0.349 -> 0.130 | Improved |
| replace RayDF with RayletDF in domain | 3D Gaussians input · ARKitScenes | ADE (m): 0.183 -> 0.115 | 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.
- Pointersect — Pointersect: Neural Rendering With Cloud-Ray Intersection (CVPR 2023) · better on 6 of 6 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
- GaussianUDF: Inferring Unsigned Distance Functions through 3D Gaussian Splatting (CVPR 2025 · similarity 0.86)
- 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes (CVPR 2025 · similarity 0.85)
- Geometry Field Splatting with Gaussian Surfels (CVPR 2025 · similarity 0.85)
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 2024 · similarity 0.85)
- Sensing Surface Patches in Volume Rendering for Inferring Signed Distance Functions (AAAI 2025 · similarity 0.84)
- Parametric Point Cloud Completion for Polygonal Surface Reconstruction (CVPR 2025 · similarity 0.84)
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.
