Research map / Point clouds / Shape and surface completion
Shape Reconstruction by Learning Differentiable Surface Representations
Jan Bednarik, Shaifali Parashar, Erhan Gundogdu, Mathieu Salzmann, Pascal Fua
Check 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 deformation regularization (ablation) | FoldingNet shape completion · FTS | mean angular error: 24.38 -> 5.94 | Improved |
| Add deformation regularization (ablation) | FoldingNet shape completion · FTS | collapsed patches: 9 -> 0 | Improved |
| Add deformation regularization (ablation) | FoldingNet shape completion · FTS | CHD: 0.14 -> 0.11 | Improved |
| Add deformation regularization (ablation) | FoldingNet single-view reconstruction · TDS cloth and T-Shirt, 2–15 patches | mean angular error: 30.32 -> 20.90; 48.32 -> 20.35; 21.95 -> 20.60; 36.29 -> 20.44; 28.77 -> 21.08; 22.95 -> 20.77; 25.67 -> 20.17; 23.51 -> | Improved |
| Add deformation regularization (ablation) | FoldingNet single-view reconstruction · TDS cloth and T-Shirt, 2–15 patches | CHD: 0.36 -> 0.33; 0.48 -> 0.51; 0.35 -> 0.32; 0.46 -> 0.49; 0.37 -> 0.39; 0.41 -> 0.42; 0.33 -> 0.41; 0.41 -> 0.41 | Mixed |
| Add differential-property and overlap losses | AtlasNet point-cloud autoencoding · ShapeNet airplane, chair, car, couch, cellphone | mean angular error: 21.26 -> 17.90; 24.49 -> 23.06; 18.08 -> 17.75; 16.83 -> 14.90; 10.29 -> 9.64 | Improved |
| Add differential-property and overlap losses | AtlasNet point-cloud autoencoding · ShapeNet airplane, chair, car, couch, cellphone | CHD: 1.07 -> 1.08; 2.79 -> 2.82; 4.68 -> 3.34; 2.10 -> 2.21; 1.80 -> 1.82 | Mixed |
| Add differential-property and overlap losses | AtlasNet single-view reconstruction · TDS cloth and T-Shirt | mean angular error: 47.42 -> 20.06; 42.12 -> 20.52 | Improved |
| Add differential-property and overlap losses | AtlasNet single-view reconstruction · TDS cloth and T-Shirt | CHD: 0.26 -> 0.28; 0.35 -> 0.31 | Mixed |
| Disable all deformation-loss terms (ablation) | FoldingNet single-view reconstruction · TDS cloth | patch collapse and distortion | Hurt |
| Add the skew term to E and G regularization (ablation) | FoldingNet single-view reconstruction · TDS cloth | patch skew | Improved |
| Omit the stretch term while retaining E, G, and skew terms (ablation) | FoldingNet single-view reconstruction · TDS cloth | patch stretching | Hurt |
Closest papers by meaning
- Shape As Points: A Differentiable Poisson Solver (NeurIPS 2021 · similarity 0.90)
- PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations (ECCV 2020 · similarity 0.89)
- DeepMesh: Differentiable Iso-Surface Extraction (TPAMI 2024 · similarity 0.88)
- SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks (CVPR 2017 · similarity 0.88)
- Coupling Explicit and Implicit Surface Representations for Generative 3D Modeling (ECCV 2020 · similarity 0.88)
- Diffeomorphic Mesh Deformation via Efficient Optimal Transport for Cortical Surface Reconstruction (ICLR 2024 · similarity 0.87)
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.
