Research map / Point clouds / Shape and surface completion
Unsupervised Deep Shape Descriptor With Point Distribution Learning
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 the proposed unsupervised descriptor instead of 3D-GAN features | ModelNet40 classification · ModelNet40 | accuracy: 83.3% -> 84.7% | Helped |
| use multi-scale descriptor fusion instead of a single σ = 0.08 descriptor (ablation) | the proposed shape descriptor · ShapeNet test set | accuracy: 94.3% -> 96.2% | Helped |
| use a single σ value of 0.04 instead of multi-scale descriptor fusion (ablation) | the proposed shape descriptor · ShapeNet test set | accuracy: 96.2% -> 91.4% | Hurt |
| use a single σ value of 0.12 instead of multi-scale descriptor fusion (ablation) | the proposed shape descriptor · ShapeNet test set | accuracy: 96.2% -> 92.9% | Hurt |
| apply rotations up to 20 degrees | the proposed shape descriptor · ShapeNet rotated test shapes | accuracy: above 93.0% | No clear effect |
| apply 45-degree rotations along each axis | the proposed shape descriptor · ShapeNet rotated test shapes | accuracy: drops by 2.3% | Hurt |
| add Gaussian noise with standard deviation less than 0.04 | the proposed shape descriptor · ShapeNet noisy test shapes | accuracy | No clear effect |
Closest papers by meaning
- Learning Descriptor Networks for 3D Shape Synthesis and Analysis (CVPR 2018 · similarity 0.88)
- Unsupervised 3D Learning for Shape Analysis via Multiresolution Instance Discrimination (AAAI 2021 · similarity 0.88)
- Unsupervised Learning of Shape and Pose with Differentiable Point Clouds (NeurIPS 2018 · similarity 0.87)
- Deep Shells: Unsupervised Shape Correspondence with Optimal Transport (NeurIPS 2020 · similarity 0.87)
- PPFNet: Global Context Aware Local Features for Robust 3D Point Matching (CVPR 2018 · similarity 0.87)
- PointGrid: A Deep Network for 3D Shape Understanding (CVPR 2018 · 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.