Research map / Point clouds / Shape and surface completion
3D Shape Generation and Completion Through Point-Voxel Diffusion
Linqi Zhou, Yilun Du, Jiajun Wu
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 |
|---|---|---|---|
| use point-voxel representation instead of pure voxel diffusion (ablation) | PVD · ShapeNet Airplane generation | 1-NN accuracy (CD): 99.75 -> 73.82 | Improved |
| use point-voxel representation instead of pure voxel diffusion (ablation) | PVD · ShapeNet Chair generation | 1-NN accuracy (CD): 97.12 -> 56.26 | Improved |
| use point-voxel representation instead of pure voxel diffusion (ablation) | PVD · ShapeNet Car generation | 1-NN accuracy (EMD): 96.83 -> 53.83 | Improved |
| use point-voxel CNN backbone instead of pure point diffusion with PointNet++ (ablation) | PVD · ShapeNet generation | visual quality | Improved |
| use PVD instead of PointFlow | PointFlow · ShapeNet Chair generation | 1-NN accuracy (CD): 62.84 -> 56.26 | Improved |
| use PVD instead of PointFlow | PointFlow · ShapeNet Car generation | 1-NN accuracy (EMD): 56.25 -> 53.83 | Improved |
| use PVD instead of PointFlow for completion | PointFlow · ShapeNet Chair completion | CD x10^3 and EMD x10^2: CD 2.707 -> 3.211; EMD 3.649 -> 2.939 | Mixed |
| use PVD instead of PointFlow for completion | PointFlow · ShapeNet Airplane completion | CD x10^3 and EMD x10^2: CD 0.4030 -> 0.4415; EMD 1.180 -> 1.030 | Mixed |
| use PVD instead of DPF-Net for completion | DPF-Net · ShapeNet Car completion | CD x10^3 and EMD x10^2: CD 1.396 -> 1.774; EMD 2.318 -> 2.146 | Mixed |
| use PVD instead of cGAN for multi-modal completion | cGAN · PartNet multi-modal completion (avg) | TMD x10^2 (diversity): 1.89 -> 3.18 | Improved |
| use PVD instead of cGAN for multi-modal completion | cGAN · PartNet multi-modal completion (avg) | MMD x10^3 (quality): 1.77 -> 1.43 | Improved |
| use PVD instead of KNN-latent for multi-modal completion | KNN-latent · PartNet multi-modal completion (avg) | MMD x10^3 (quality): 1.57 -> 1.43 | No clear effect |
Closest papers by meaning
- Probabilistic Implicit Scene Completion (ICLR 2022 · similarity 0.91)
- Diffusion-Based Signed Distance Fields for 3D Shape Generation (CVPR 2023 · similarity 0.91)
- GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors (CVPR 2024 · similarity 0.90)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (CVPR 2025 · similarity 0.89)
- Sin3DM: Learning a Diffusion Model from a Single 3D Textured Shape (ICLR 2024 · similarity 0.88)
- Neural Point Cloud Diffusion for Disentangled 3D Shape and Appearance Generation (CVPR 2024 · 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.
