Research map / Point clouds / Point cloud classification
Benchmarking and Analyzing Robust Point Cloud Recognition: Bag of Tricks for Defending Adversarial Examples
Qiufan Ji, Lin Wang, Cong Shi, Shengshan Hu, Yingying Chen, Lichao Sun
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 statistical outlier removal (SOR) pre-processing | PointNet, no defense · ModelNet40, AOF attack | accuracy: 54.54 -> 68.48 | Improved |
| add IF-Defense (ConvONet reconstruction) | PointNet, no defense · ModelNet40, SIA attack | accuracy: 31.40 -> 43.76 | Improved |
| add hybrid training (add, remove, shift attacks) as augmentation | PointNet, no defense · ModelNet40, SIA attack | accuracy: 31.40 -> 76.26 | Improved |
| combine SOR, hybrid training and ConvONet into one framework (ablation) | defense pipeline · ModelNet40, average over 11 attacks and 8 models | average defense accuracy: 79.35 -> 83.45 | Improved |
| remove hybrid training from the framework (SOR plus ConvONet only) | full defense framework · ModelNet40, AOF/GeoA3/SIA | accuracy | Hurt |
| apply hybrid training defense against KNN and L3A attacks | hybrid training on PointNet · ModelNet40, KNN attack | accuracy: KNN 64.71 vs No Defense 45.10 still higher, but far below SOR-free framework 85.53 | Hurt |
| apply SOR defense on PointConv clean inputs | PointConv · ModelNet40, clean | clean accuracy: 88.65 -> 72.12 | Hurt |
| use models with advanced grouping (CurveNet, GDANet, RPC) | victim models · ModelNet40, various attacks | adversarial robustness | Improved |
Closest papers by meaning
- Minimal Adversarial Examples for Deep Learning on 3D Point Clouds (ICCV 2021 · similarity 0.94)
- Improving Adversarial Robustness of 3D Point Cloud Classification Models (ECCV 2022 · similarity 0.94)
- A Critical Revisit of Adversarial Robustness in 3D Point Cloud Recognition with Diffusion-Driven Purification (ICML 2023 · similarity 0.93)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.90)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (AAAI 2024 · similarity 0.90)
- CausalPC: Improving the Robustness of Point Cloud Classification by Causal Effect Identification (CVPR 2024 · similarity 0.90)
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.
