Research map / 3D vision / Face and shape
Efficient Dense Point Cloud Object Reconstruction using Deformation Vector Fields
Kejie Li, Trung Pham, Huangying Zhan, Ian Reid
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 distance field multi-view consistency loss (ablation) | deformed depth map network · ShapeNet | overall Chamfer Distance: 3.102 -> 2.987 | Improved |
| add binary mask multi-view consistency loss instead of distance field (ablation) | depth-mask network of Lin et al. · ShapeNet | overall Chamfer Distance: 3.240 -> 3.531 | Hurt |
| replace binary mask multi-view loss with distance field loss (ablation) | deformed depth map network · ShapeNet | prediction points: 25401 -> 98304 | Improved |
| replace binary mask multi-view loss with distance field loss (ablation) | deformed depth map network · ShapeNet | overall Chamfer Distance: 3.531 -> 2.987 | Improved |
| replace depth-mask representation with deformed depth map (ablation) | single-view CNN · ShapeNet (plane, car, chair) | Chamfer Distance | Improved |
| raise foreground threshold of depth-mask baseline (ablation) | depth-mask representation · ShapeNet (plane, car, chair) | pred to gt vs gt to pred distance | Mixed |
| use deformed depth map instead of PSGN point cloud | single-view reconstruction · ShapeNet 13 classes | Chamfer Distance: 11 out of 13 categories | Improved |
| use deformed depth map instead of PSGN on plane | single-view reconstruction · ShapeNet plane | Chamfer Distance: 2.582 -> 2.66 | Hurt |
| use deformed depth map instead of PSGN on gun | single-view reconstruction · ShapeNet gun | Chamfer Distance: 2.949 -> 3.259 | Hurt |
| densify PSGN from 1024 to 98304 points by interpolation | PSGN · ShapeNet (5 categories) | mean Chamfer Distance | No clear effect |
| use deformed depth map instead of HSP octree voxels | single-view reconstruction · ShapeNet chair | Chamfer Distance: 4.716 -> 3.731 | Improved |
| use deformed depth map instead of HSP on plane | single-view reconstruction · ShapeNet plane | Chamfer Distance: 3.878 -> 2.660 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MarrNet — MarrNet: 3D Shape Reconstruction via 2.5D Sketches (NeurIPS 2017) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Point Set Generation Network for 3D Object Reconstruction From a Single Image (CVPR 2017 · similarity 0.90)
- Learning Efficient Point Cloud Generation for Dense 3D Object Reconstruction (AAAI 2018 · similarity 0.90)
- MVPNet: Multi-View Point Regression Networks for 3D Object Reconstruction from A Single Image (AAAI 2019 · similarity 0.89)
- GraphX-Convolution for Point Cloud Deformation in 2D-to-3D Conversion (ICCV 2019 · similarity 0.89)
- From Image Collections to Point Clouds With Self-Supervised Shape and Pose Networks (CVPR 2020 · similarity 0.88)
- Depth-Guided Robust and Fast Point Cloud Fusion NeRF for Sparse Input Views (AAAI 2024 · similarity 0.86)
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.
