Research map / 3D vision / Face and shape
Learning elementary structures for 3D shape generation and matching
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Learn shared point-translation elementary structures instead of using AtlasNet's fixed square patches | ShapeNet Core reconstruction with MLP adjustment · ShapeNet Core multi-category All | Chamfer distance (×10^-3): 1.45 -> 1.22 | Helped |
| Learn shared elementary structures instead of using the fixed 3D-CODED template | Human reconstruction · SURREAL test | Chamfer loss: 1.33 -> 1.11 | Helped |
| Use continuous patch-deformation elementary structures rather than the fixed 3D-CODED template | Human shape matching · FAUST inter challenge | Correspondence error: 2.96 -> 2.76 | Helped |
| Use MLP adjustment rather than linear adjustment (ablation) | Learned elementary structures · ShapeNet Core multi-category All | Chamfer distance (×10^-3): 2.11 -> 1.22 | Helped |
| Use point translation rather than patch deformation (ablation) | Learned elementary structures · ShapeNet Core multi-category All | Chamfer distance (×10^-3): 1.43 -> 1.22 | Helped |
| Use point translation rather than patch deformation (ablation) | Human reconstruction and matching · SURREAL test and FAUST inter challenge | Chamfer loss and correspondence error: 1.17 -> 1.11; 2.76 -> 3.05 | Mixed |
| Increase elementary-structure dimensionality from 3D to 10D (ablation) | ShapeNet Core reconstruction with MLP adjustment · ShapeNet Core multi-category All | Chamfer distance (×10^-3): 1.22 -> 1.21 | Helped |
| Increase elementary-structure dimensionality from 3D to 10D (ablation) | Human reconstruction and matching · SURREAL test and FAUST inter challenge | Chamfer loss and correspondence error: 1.11 -> 1.01; 3.05 -> 2.85 | Helped |
| Increase elementary-structure dimensionality from 3D to 10D (ablation) | Human reconstruction and matching with patch deformation · SURREAL test and FAUST inter challenge | Chamfer loss and correspondence error: 1.17 -> 1.01; 2.76 -> 2.77 | Mixed |
| Use 2D rather than 3D elementary structures (ablation) | ShapeNet Core reconstruction with MLP adjustment · ShapeNet Core multi-category All | Chamfer distance (×10^-3): 1.22 -> 1.28 | Hurt |
| Use 2D rather than 3D elementary structures (ablation) | Human reconstruction and matching · SURREAL test and FAUST inter challenge | Chamfer loss and correspondence error: 1.11 -> 1.55; 3.05 -> 3.28 | Hurt |
Closest papers by meaning
- A Papier-Mâché Approach to Learning 3D Surface Generation (CVPR 2018 · similarity 0.88)
- Unsupervised Learning of Dense Shape Correspondence (CVPR 2019 · similarity 0.88)
- SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks (CVPR 2017 · similarity 0.88)
- Unsupervised Learning of Intrinsic Structural Representation Points (CVPR 2020 · similarity 0.88)
- Learning Shape Abstractions by Assembling Volumetric Primitives (CVPR 2017 · similarity 0.88)
- Geometry-Aware Network for Non-Rigid Shape Prediction From a Single View (CVPR 2018 · 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.