Research map / 3D vision / Face and shape
MeshUDF: Fast and Differentiable Meshing of Unsigned Distance Field Networks
Benoît Guillard, Federico Stella, Pascal Fua
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 border gradient term to normals-only gradients (ablation) | MeshUDF point cloud fitting · Sparse point cloud fitting | CHD: 3.75 -> 3.54 | Improved |
| add border gradient term (ablation) | MeshUDF silhouette fitting · Silhouette fitting | CHD: 10.45 -> 9.68 | Improved |
| add border gradient term (ablation) | MeshUDF silhouette fitting · Silhouette fitting | IC: 78.84 -> 79.90 | Improved |
| use breadth-first exploration with voting instead of pseudo-sign (PSD) (ablation) | MeshUDF · MGN garments, resolution 512 | excess holes: 11526 -> 478 | Improved |
| use breadth-first exploration with voting instead of PSD (ablation) | MeshUDF · MGN garments, resolution 64 | CHD: 1.63 -> 1.66 | Hurt |
| use breadth-first exploration with voting instead of PSD (ablation) | MeshUDF · MGN garments, resolution 512 | time: 105s -> 69s | Improved |
| mesh the UDF with MeshUDF instead of inflation | UDF meshing · MGN garments | CHD: 3.00 -> 1.51 | Improved |
| mesh the UDF with MeshUDF instead of inflation | UDF meshing · ShapeNet cars | NC: 73.19 -> 70.38 | Hurt |
| mesh the UDF with MeshUDF instead of ball pivoting | UDF meshing · MGN garments | time: 16.5s + 3000s -> 1.2 sec. | Improved |
| optimize mesh-based loss L_PC,mesh instead of L_PC,UDF in implicit domain | sparse point cloud fitting · Sparse point cloud fitting | CHD: 4.54 -> 3.54 | Improved |
| optimize L_silh,mesh instead of L_silh,UDF | silhouette fitting · Silhouette fitting | CHD: 12.74 -> 9.68 | Improved |
| start from random latent code instead of same-class code | L_PC,mesh fitting · Sparse point cloud fitting | CHD: 3.54 -> 3.64 | No clear effect |
Closest papers by meaning
- High Resolution UDF Meshing via Iterative Networks (NeurIPS 2025 · similarity 0.90)
- MIND: Material Interface Generation from UDFs for Non-Manifold Surface Reconstruction (NeurIPS 2025 · similarity 0.89)
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.88)
- Unsigned Orthogonal Distance Fields: An Accurate Neural Implicit Representation for Diverse 3D Shapes (CVPR 2024 · similarity 0.85)
- UDiFF: Generating Conditional Unsigned Distance Fields with Optimal Wavelet Diffusion (CVPR 2024 · similarity 0.84)
- Details Enhancement in Unsigned Distance Field Learning for High-fidelity 3D Surface Reconstruction (AAAI 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.
