Research map / Point clouds / Point cloud classification
PointCert: Point Cloud Classification With Deterministic Certified Robustness Guarantees
Jinghuai Zhang, Jinyuan Jia, Hongbin Liu, Neil Zhenqiang Gong
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 deterministic disjoint hash-based sub-point clouds and majority voting instead of randomized smoothing or PointGuard | Point cloud classification on ModelNet40 · ModelNet40 | Certified accuracy@50: 79% vs. $0$ | Improved |
| Train the base classifier on sub-point clouds rather than standard point clouds (ablation) | PointCert Scenario II vs. Scenario I · ModelNet40 and ScanObjectNN variants | certified accuracy | Improved |
| Prepend a Point Completion Network before the standard-trained classifier (ablation) | PointCert Scenario III vs. Scenario I · ModelNet40 and ScanObjectNN variants | certified accuracy | Improved |
| Add the classifier-oriented loss term to PCN training (lambda > 0 rather than lambda = 0) (ablation) | PointCert Scenario III · ModelNet40 and ScanObjectNN variants | certified accuracy | Improved |
| Vary the number of hash-based sub-point clouds; smaller m raises clean accuracy but weakens robustness (ablation) | PointCert Scenario II · ModelNet40 | certified accuracy | Mixed |
| Use MD5 rather than the mean-based hash function (ablation) | PointCert Scenario II · ModelNet40 | certified accuracy | Improved |
| Use different base point cloud classifiers (ablation) | PointCert Scenario II · ModelNet40 | certified accuracy | No clear effect |
| Increase the PCN classification-loss weight lambda; certified accuracy rises and then falls (ablation) | PointCert Scenario III · ModelNet40 | certified accuracy | Mixed |
| Use a pretrained PCN rather than training the PCN from scratch (ablation) | PointCert Scenario III · ModelNet40 | certified accuracy | Improved |
| Increase the fraction of customer point clouds with labels (ablation) | PointCert Scenario III · ModelNet40 | certified accuracy | Improved |
| Use knowledge-distilled student model in the black-box setting rather than the teacher model in the white-box setting (ablation) | PointCert Scenario III · ModelNet40 | certified accuracy | No clear effect |
| Use PointGuard instead of PointCert under empirical point deletion attacks | Point cloud classification · ModelNet40 and ScanObjectNN variants | empirical accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PointGuard — PointGuard: Provably Robust 3D Point Cloud Classification (CVPR 2021) · 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
- PointGuard: Provably Robust 3D Point Cloud Classification (CVPR 2021 · similarity 0.91)
- Certified L2-Norm Robustness of 3D Point Cloud Recognition in the Frequency Domain (AAAI 2026 · similarity 0.91)
- PointCVaR: Risk-Optimized Outlier Removal for Robust 3D Point Cloud Classification (AAAI 2024 · similarity 0.88)
- Improving Robustness of 3D Point Cloud Recognition from a Fourier Perspective (NeurIPS 2024 · similarity 0.87)
- CausalPC: Improving the Robustness of Point Cloud Classification by Causal Effect Identification (CVPR 2024 · similarity 0.87)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.85)
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.
