Research map / Point clouds / Shape and surface completion
PointGMM: A Neural GMM Network for Point Clouds
Check 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 |
|---|---|---|---|
| use PointGMM for rigid registration instead of the strongest listed baseline | PointGMM registration · Partial chair, car, and airplane point clouds across the reported rotation and coverage conditions | MSE: chair: 0.0430 -> 0.0226; car: 0.0399 -> 0.0246; airplane: 0.0632 -> 0.0312 (30°, 50–80% coverage) | Helped |
| use PointGMM instead of the strongest listed baseline under large rotations | PointGMM registration · Chair, car, and airplane; 180° max rotation, both coverage ranges | MSE: chair: 0.2578 -> 0.0232; 0.2817 -> 0.0574; car: 0.2134 -> 0.0290; 0.2350 -> 0.0702; airplane: 0.1979 -> 0.0350; 0.2293 - | Helped |
| add a PointGMM decoder to the transformation encoder (ablation) | PointGMM registration · Chair, car, and airplane decoder ablation | MSE: chair: 0.0612 -> 0.0382; car: 0.1514 -> 0.0447; airplane: 0.1688 -> 0.0447 | Helped |
| replace the PointGMM decoder with a PointNet auto-encoder point decoder (ablation) | PointGMM registration · Chair, car, and airplane decoder ablation | MSE: chair: 0.0382 -> 0.1185; car: 0.0447 -> 0.2073; airplane: 0.0447 -> 0.1817 | Hurt |
| use a hierarchical GMM rather than a vanilla GMM (ablation) | PointGMM shape generation · ShapeNet chair | log-likelihood | Helped |
| increase Gaussian count in the hierarchical model (ablation) | PointGMM shape generation · ShapeNet chair | log-likelihood | Helped |
| increase Gaussian count in the vanilla GMM model (ablation) | PointGMM shape generation · ShapeNet chair | log-likelihood | Hurt |
| add sibling self-attention to the hierarchical decoder (ablation) | PointGMM shape generation · ShapeNet chair, especially shallower models with fewer Gaussians | log-likelihood | Helped |
Closest papers by meaning
- Accelerated Generative Models for 3D Point Cloud Data (CVPR 2016 · similarity 0.89)
- Learning Representations and Generative Models for 3D Point Clouds (ICML 2018 · similarity 0.88)
- PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priors (CVPR 2025 · similarity 0.88)
- A Linear Time and Space Local Point Cloud Geometry Encoder via Vectorized Kernel Mixture (VecKM) (ICML 2024 · similarity 0.87)
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.87)
- Rectified Point Flow: Generic Point Cloud Pose Estimation (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.