Research map / Point clouds / Point cloud classification
Constraint-Aware Feature Learning for Parametric Point Cloud
Xi Cheng, Ruiqi Lei, Di Huang, Zhichao Liao, Fengyuan Piao, Yan Chen, Pingfa Feng, Long Zeng
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 MAD component (ablation) | CstNet · Param20K | instance accuracy: 86.32 -> 89.94 | Improved |
| add Adjacency component (ablation) | CstNet · Param20K | instance accuracy: 88.14 -> 89.94 | Improved |
| add Primitive Type component (ablation) | CstNet · Param20K | instance accuracy: 88.68 -> 89.94 | Improved |
| add all three constraint components (ablation) | CstNet · Param20K | instance accuracy: 83.99 -> 89.94 | Improved |
| use SurfaceKNN instead of KNN (ablation) | CstNet · Param20K | instance accuracy: 89.03 -> 89.94 | Improved |
| use SurfaceKNN when Primitive Type disabled (ablation) | CstNet · Param20K | instance accuracy: 88.73 -> 88.68 | No clear effect |
| use predicted instead of label constraints | CstNet · Param20K | instance accuracy: 91.80 -> 89.94 | Hurt |
| use Stage 2 backbone of CstNet vs DGCNN with predicted constraints (ablation) | CstNet · Param20K | instance accuracy: 86.81 -> 89.94 | Improved |
| add constraint-aware features to PointNet++ for prism-cuboid classification (ablation) | PointNet++ · Prism-Cuboid 50 to 87 degrees | accuracy | Improved |
| add constraints on 89 degree prisms | CstNet · Prism-Cuboid 89 degrees | accuracy | No clear effect |
| CstPnt vs ParSeNet on unseen Param20K Primitive Type prediction | CstPnt · Param20K (unseen) | primitive type accuracy: 59.75 -> 86.52 | Improved |
| CstNet vs PTMamba classification | CstNet · Param20K | instance accuracy: 86.45 -> 89.94 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- HPNet — HPNet: Deep Primitive Segmentation Using Hybrid Representations (ICCV 2021) · better on 7 of 7 shared results
- PointConv — PointConv: Deep Convolutional Networks on 3D Point Clouds (CVPR 2019) · better on 1 of 1 shared result
- PointCNN — PointCNN: Convolution On X-Transformed Points (NeurIPS 2018) · better on 1 of 1 shared result
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
- PIE-NET: Parametric Inference of Point Cloud Edges (NeurIPS 2020 · similarity 0.87)
- MiCADangelo: Fine-Grained Reconstruction of Constrained CAD Models from 3D Scans (NeurIPS 2025 · similarity 0.85)
- CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention (CVPR 2024 · similarity 0.85)
- Rethinking Point Cloud Data Augmentation: Topologically Consistent Deformation (ICML 2025 · similarity 0.84)
- CADCrafter: Generating Computer-Aided Design Models from Unconstrained Images (CVPR 2025 · similarity 0.84)
- Point2CAD: Reverse Engineering CAD Models from 3D Point Clouds (CVPR 2024 · similarity 0.84)
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.
