Research map / Point clouds / Shape and surface completion
Learning a More Continuous Zero Level Set in Unsigned Distance Fields through Level Set Projection
Junsheng Zhou, Baorui Ma, Shujuan Li, Yu-Shen Liu, 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 |
|---|---|---|---|
| add the three zero level set constraints (projection, distance, orthogonal) (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.126 -> 0.091 | Improved |
| add level set projection loss (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.107 -> 0.091 | Improved |
| add surface unsigned distance loss (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.101 -> 0.091 | Improved |
| add gradient-surface orthogonal loss (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.096 -> 0.091 | Improved |
| add adaptive per-point weights (ablation) | LevelSetUDF · ShapeNet Cars subset | Normal Consistency: 84.2 -> 84.4 | Improved |
| use projection loss on the zero level set instead of GenSDF-style optimization (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.101 -> 0.091 | Improved |
| apply projection loss on the zero level set rather than a random non-zero level set (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.098 -> 0.091 | Improved |
| increase projection loss weight from 0.002 to 0.1 (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.091 -> 0.359 | Hurt |
| increase orthogonal loss weight from 0.01 to 1 (ablation) | LevelSetUDF · ShapeNet Cars subset | L2 Chamfer (x10^4): 0.091 -> 2.55 | Hurt |
| add zero level set constraints for unsupervised normal estimation (ablation) | LevelSetUDF · PCPNet (average) | angle RMSE: 7.61 -> 7.20 | Improved |
| add zero level set constraints for point cloud upsampling (ablation) | LevelSetUDF · PU-GAN | HD: 2.16 -> 1.55 | Improved |
| vary adaptive weight lambda from 10 to 1000 (ablation) | LevelSetUDF · ShapeNet Cars subset | Normal Consistency: 84.4 -> 83.2 | Hurt |
Closest papers by meaning
- A Lightweight UDF Learning Framework for 3D Reconstruction Based on Local Shape Functions (CVPR 2025 · similarity 0.90)
- Details Enhancement in Unsigned Distance Field Learning for High-fidelity 3D Surface Reconstruction (AAAI 2025 · similarity 0.89)
- Small Steps and Level Sets: Fitting Neural Surface Models with Point Guidance (CVPR 2024 · similarity 0.89)
- High Resolution UDF Meshing via Iterative Networks (NeurIPS 2025 · similarity 0.88)
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.88)
- Learning Bijective Surface Parameterization for Inferring Signed Distance Functions from Sparse Point Clouds with Grid Deformation (CVPR 2025 · similarity 0.88)
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.
