Research map / Point clouds / Point cloud classification
HHA: Hierarchical Hyperbolic Constraints for Imperceptible Point Cloud Attacks
Keke Tang, Yu Liao, Weilong Peng, Xiaofei Wang, Daizong Liu, Zhongyun Hua, Peican Zhu, Zhihong Tian
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 hierarchical hyperbolic regularization (full HHA) vs best baseline LBC | HHA · ModelNet40 PointNet | CD (1e-4): 0.897 -> 0.603 | Improved |
| add hierarchical hyperbolic regularization vs LBC | HHA · ScanObjectNN PointNet | CD (1e-4): 0.879 -> 0.474 | Improved |
| add semantic regularization to geometric-only (ablation) | HHA · ModelNet40 PointNet | CD (1e-4): 0.686 -> 0.603 | Improved |
| add geometric regularization to semantic-only (ablation) | HHA · ModelNet40 PointNet | CD (1e-4): 0.621 -> 0.603 | Improved |
| add semantic regularization to geometric-only (ablation) | HHA · ScanObjectNN PointNet | CD (1e-4): 0.510 -> 0.474 | Improved |
| use hyperbolic distance instead of Euclidean in component regularization (ablation) | HHA · ModelNet40 PointNet | CD (1e-4): 1.034 -> 0.603 | Improved |
| use hyperbolic distance instead of Euclidean in component regularization (ablation) | HHA · ModelNet40 PointNet | HD (1e-2): 0.511 -> 0.279 | Improved |
| increase curvature c from 0.01 to 1.0 (ablation) | HHA · ModelNet40 PointNet | CD (1e-4): 0.805 -> 0.603 | Improved |
| increase curvature c from 1.0 to 1.5 (ablation) | HHA · ModelNet40 PointNet | CD (1e-4): 0.603 -> 0.757 | Hurt |
| use Poincare ball model instead of Lorentz model (ablation) | HHA · ModelNet40 PointNet | CD (1e-4): 2.108 -> 0.603 | Improved |
| use stronger P3-SAM segmenter instead of PartSLIP (ablation) | HHA · ScanObjectNN PointNet | CD (1e-4): 0.474 -> 0.469 | No clear effect |
| attack with HHA under IF-Defense instead of LBC | HHA · ModelNet40 PointNet | ASR (%): 45.2 -> 75.7 | Improved |
Closest papers by meaning
- Manifold Constraints for Imperceptible Adversarial Attacks on Point Clouds (AAAI 2024 · similarity 0.90)
- FLAT: Flux-aware Imperceptible Adversarial Attacks on 3D Point Clouds (ECCV 2024 · similarity 0.90)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.89)
- Imperceptible 3D Point Cloud Attacks on Lattice-based Barycentric Coordinates (AAAI 2025 · similarity 0.88)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.88)
- Explicitly Perceiving and Preserving the Local Geometric Structures for 3D Point Cloud Attack (AAAI 2024 · similarity 0.88)
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.
