Research map / Point clouds / Shape and surface completion
Learning Consistency-Aware Unsigned Distance Functions Progressively from Raw Point Clouds
Junsheng Zhou, Baorui Ma, Yu-Shen Liu, Yi Fang, Zhizhong Han
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 |
|---|---|---|---|
| replace the two-stage progressive loss with the Neural-Pull loss (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1112 -> 0.2381 | Hurt |
| replace |x| with 1-exp(-x) at the last layer (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1112 -> 0.1218 | Hurt |
| train second stage from scratch instead of end-to-end (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1112 -> 0.1497 | Hurt |
| increase training stages from 1 to 2 (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1218 -> 0.1112 | Improved |
| increase training stages from 2 to 3 or 4 (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1112 -> 0.1107 | No clear effect |
| low confidence range 0.9 sigma instead of 1.1 sigma (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1112 -> 0.1133 | Hurt |
| low confidence range 1.2 sigma instead of 1.1 sigma (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1112 -> 0.1131 | Hurt |
| add mesh refinement using UDF values at 64^3 grid (ablation) | Ours · ShapeNet Car subset | L2CD: 0.4169 -> 0.1606 | Improved |
| add mesh refinement at 256^3 grid (ablation) | Ours · ShapeNet Car subset | L2CD: 0.1294 -> 0.1112 | Improved |
| increase grid resolution from 64^3 to 320^3 (ablation) | Ours · ShapeNet Car subset | L2CD and time: 0.1606 -> 0.1105; 3.0 s -> 307.6 s | Mixed |
| apply our gradient-based reconstruction to NDF instead of BPA | NDF · ShapeNet cars | Chamfer-L2 mean: 0.202 -> 0.160 | Improved |
| use our UDF method instead of NDF | NDF · MGD | Chamfer-L2: 0.658 -> 0.117 | 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.
- Neural-Pull — Neural-Pull: Learning Signed Distance Function from Point clouds by Learning to Pull Space onto Surface (ICML 2021) · better on 3 of 3 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
- CAP-UDF: Learning Unsigned Distance Functions Progressively From Raw Point Clouds With Consistency-Aware Field Optimization (TPAMI 2024 · similarity 0.97)
- A Lightweight UDF Learning Framework for 3D Reconstruction Based on Local Shape Functions (CVPR 2025 · similarity 0.93)
- Details Enhancement in Unsigned Distance Field Learning for High-fidelity 3D Surface Reconstruction (AAAI 2025 · similarity 0.91)
- Learning Bijective Surface Parameterization for Inferring Signed Distance Functions from Sparse Point Clouds with Grid Deformation (CVPR 2025 · similarity 0.90)
- Inferring Neural Signed Distance Functions by Overfitting on Single Noisy Point Clouds through Finetuning Data-Driven based Priors (NeurIPS 2024 · similarity 0.90)
- MultiPull: Detailing Signed Distance Functions by Pulling Multi-Level Queries at Multi-Step (NeurIPS 2024 · 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.
