Research map / Point clouds / Point cloud classification
Verifiability and Predictability: Interpreting Utilities of Network Architectures for Point Cloud Processing
Wen Shen, Zhihua Wei, Shikun Huang, Binbin Zhang, Panyue Chen, Ping Zhao, Quanshi Zhang
Check 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 |
|---|---|---|---|
| Keep Architecture 1, which reweights features using local density (ablation) | PointConv · ModelNet40, ShapeNet, and 3D MNIST | adversarial robustness: 0.249; 0.136; 0.207 | Improved |
| Keep Architecture 2, which reweights features using local coordinates (ablation) | PointConv · ModelNet40, ShapeNet, and 3D MNIST | rotation non-robustness: 0.191; 0.898; 1.369 | Improved |
| Increase Point2Sequence scales from 3 to 4 (ablation) | Point2Sequence · ModelNet40, ShapeNet, and 3D MNIST | adversarial robustness: 0.005; 0.006; -0.011 | Mixed |
| Increase Point2Sequence scales from 2 to 4 (ablation) | Point2Sequence · ModelNet40, ShapeNet, and 3D MNIST | adversarial robustness: 0.013; 0.032; 0.008 | Improved |
| Increase Point2Sequence scales from 3 to 4 (ablation) | Point2Sequence · ModelNet40, ShapeNet, and 3D MNIST | neighborhood inconsistency: 0.148; 0.177; 0.245 | Improved |
| Increase Point2Sequence scales from 2 to 4 (ablation) | Point2Sequence · ModelNet40, ShapeNet, and 3D MNIST | neighborhood inconsistency: -0.036; 0.132; 0.334 | Mixed |
| Keep Architecture 4, which encodes orientation-aware features (ablation) | PointSIFT · ModelNet40, ShapeNet, and 3D MNIST | rotation non-robustness: 3.343; 2.545; -0.689 | Mixed |
| Add Architecture 1 | PointNet++, Point2Sequence, and RSCNN · ModelNet40 | adversarial robustness: 0.015; 0.018; 0.028 | Improved |
| Add Architecture 1 | Point2Sequence · ShapeNet | adversarial robustness: -0.020 | Hurt |
| Add Architecture 2 | PointNet++ and Point2Sequence · ModelNet40 | rotation non-robustness: 3.248; 4.543 | Improved |
| Add Architecture 2 | RSCNN · ModelNet40 and ShapeNet | rotation non-robustness: -0.348; -1.225 | Hurt |
| Add Architecture 3 | PointNet++ and RSCNN · ModelNet40 | adversarial robustness: 0.506; 0.036 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PointConv — PointConv: Deep Convolutional Networks on 3D Point Clouds (CVPR 2019) · better on 1 of 1 shared result
- Point2Sequence — Point2Sequence: Learning the Shape Representation of 3D Point Clouds with an Attention-Based Sequence to Sequence Network (AAAI 2019) · 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
- Interpreting Representation Quality of DNNs for 3D Point Cloud Processing (NeurIPS 2021 · similarity 0.92)
- Benchmarking and Analyzing Robust Point Cloud Recognition: Bag of Tricks for Defending Adversarial Examples (ICCV 2023 · similarity 0.87)
- Improving Adversarial Robustness of 3D Point Cloud Classification Models (ECCV 2022 · similarity 0.87)
- LitePT: Lighter Yet Stronger Point Transformer (CVPR 2026 · similarity 0.87)
- CAP: Robust Point Cloud Classification via Semantic and Structural Modeling (CVPR 2023 · similarity 0.86)
- A Network Architecture for Point Cloud Classification via Automatic Depth Images Generation (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.
