Research map / Point clouds / Shape and surface completion
MVPNet: Multi-View Point Regression Networks for 3D Object Reconstruction from A Single Image
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 |
|---|---|---|---|
| add the full GeoLoss instead of using only point-wise distance loss (ablation) | MVPNet with 6 viewpoints · ShapeNet-13 | mean voxel IoU: 0.526 -> 0.671 | Helped |
| use 4 viewpoints instead of 6 (ablation) | MVPNet with GeoLoss · ShapeNet-13 | mean voxel IoU: 0.671 -> 0.665 | Hurt |
| use 8 viewpoints instead of 6 (ablation) | MVPNet with GeoLoss · ShapeNet-13 | mean voxel IoU: 0.671 -> 0.662 | Hurt |
| use 4 viewpoints instead of 6 (ablation) | MVPNet · ShapeNet-13 | mean Chamfer distance (scaled by 0.01): 0.096 -> 0.113 | Hurt |
| use 8 viewpoints instead of 6 (ablation) | MVPNet · ShapeNet-13 | mean Chamfer distance (scaled by 0.01): 0.096 -> 0.107 | Hurt |
| regress 3D coordinates rather than depth values (ablation) | MVPNet with the same architecture and GeoLoss · ShapeNet-13 | voxel IoU: about 4% lower | Helped |
| use MVPNet with GeoLoss rather than the strongest listed point-generation baseline | MVPNet · ShapeNet-13 | mean voxel IoU: 0.640 -> 0.671 | Helped |
| use MVPNet rather than the strongest listed point-generation baseline | MVPNet · ShapeNet-13 | mean Chamfer distance (scaled by 0.01): 2.084 -> 0.096 | Helped |
Closest papers by meaning
- A Point Set Generation Network for 3D Object Reconstruction From a Single Image (CVPR 2017 · similarity 0.91)
- POCO: Point Convolution for Surface Reconstruction (CVPR 2022 · similarity 0.90)
- Learning Efficient Point Cloud Generation for Dense 3D Object Reconstruction (AAAI 2018 · similarity 0.89)
- From Image Collections to Point Clouds With Self-Supervised Shape and Pose Networks (CVPR 2020 · similarity 0.89)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.88)
- Hypernetwork approach to generating point clouds (ICML 2020 · 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.