Research map / 3D vision / Face and shape
A Papier-Mâché Approach to Learning 3D Surface Generation
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 |
|---|---|---|---|
| Represent the output as learned surface patches rather than unstructured points | AtlasNet versus Points baseline · ShapeNet Core v2 auto-encoding | Chamfer distance (×10^3): 1.91 -> 1.51 | Helped |
| Represent the output as learned surface patches rather than points with predicted normals | AtlasNet versus Points baseline + normals · ShapeNet Core v2 auto-encoding | Metro distance (×10): 1.82 -> 1.41 | Helped |
| Predict normals in the point baseline (ablation) | Points baseline · ShapeNet Core v2 auto-encoding | Chamfer distance (×10^3): 1.91 -> 2.15 | Hurt |
| Use 125 patches rather than 1 patch (ablation) | AtlasNet · ShapeNet Core v2 auto-encoding | Chamfer distance (×10^3): 1.84 -> 1.51 | Helped |
| Use 125 patches rather than 1 patch (ablation) | AtlasNet · ShapeNet Core v2 auto-encoding | Metro distance (×10): 1.53 -> 1.41 | Helped |
| Add a spherical parameterization instead of a single square patch (ablation) | AtlasNet · ShapeNet Core v2 auto-encoding | Chamfer distance (×10^3): 1.84 -> 1.72 | Helped |
| Add a spherical parameterization instead of a single square patch (ablation) | AtlasNet · ShapeNet Core v2 auto-encoding | Metro distance (×10): 1.53 -> 1.52 | No clear effect |
| Use 125 rather than 25 patches (ablation) | AtlasNet · ShapeNet Core v2 auto-encoding | Metro distance (×10): 1.47 -> 1.41 | Mixed |
| Use 125 rather than 25 patches (ablation) | AtlasNet · ShapeNet Core v2 auto-encoding | Chamfer distance (×10^3): 1.56 -> 1.51 | Helped |
| Use 25 learned patches rather than HSP | AtlasNet · ShapeNet single-view reconstruction | Chamfer distance (×10^3): 11.6 -> 9.52 | Helped |
| Use 25 learned patches rather than HSP | AtlasNet · ShapeNet single-view reconstruction | Metro distance (×10): 1.49 -> 1.09 | Helped |
| Remove weight decay (ablation) | AtlasNet auto-encoder · ShapeNet Core v2 auto-encoding | Chamfer distance (×10^3): 8.57 -> 1.56 | Helped |
Closest papers by meaning
- SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks (CVPR 2017 · similarity 0.91)
- Neural Convolutional Surfaces (CVPR 2022 · similarity 0.89)
- Synthesizing 3D Shapes via Modeling Multi-View Depth Maps and Silhouettes With Deep Generative Networks (CVPR 2017 · similarity 0.88)
- Learning elementary structures for 3D shape generation and matching (NeurIPS 2019 · similarity 0.88)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (CVPR 2025 · similarity 0.87)
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction (NeurIPS 2025 · 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.