Research map / Point clouds / Point cloud classification
Improving Adversarial Robustness of 3D Point Cloud Classification Models
Guanlin Li, Guowen Xu, Han Qiu, Ruan He, Jiwei Li, Tianwei Zhang
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 CCN architecture on top of DGCNN, both trained with PointCutMix-K (ablation) | DGCNN · ModelNet40 white-box | AAUA: 74.82 -> 76.35 | Improved |
| add CCN architecture on top of DGCNN, both trained with PointCutMix-K (ablation) | DGCNN · ModelNet40 white-box | LAUA: 68.47 -> 68.30 | No clear effect |
| use CCN instead of PointNet, PointCutMix-K training (ablation) | PointNet · ModelNet40 white-box | AAUA: 72.84 -> 76.35 | Improved |
| use receptive field alpha=4 instead of alpha=8 in CCM (ablation) | CCN · ModelNet40 white-box | AAUA: 73.88 -> 76.35 | Improved |
| use receptive field alpha=4 instead of alpha=20 in CCM (ablation) | CCN · ModelNet40 white-box | AAUA: 75.62 -> 76.35 | Improved |
| replace normal training by AMS (ablation) | PointNet · ModelNet40 white-box | AAUA: 47.66 -> 77.44 | Improved |
| replace AT-BIM by AMS (T=0.5) (ablation) | PointNet · ModelNet40 white-box | AAUA: 75.54 -> 77.44 | Improved |
| lower AMS threshold T from 0.5 to 0.1 (ablation) | PointNet + AMS · ModelNet40 white-box | AAUA: 77.44 -> 71.03 | Hurt |
| replace PointCutMix-K by AMS on CCN (ablation) | CCN · ModelNet40 white-box | AAUA: 76.35 -> 84.74 | Improved |
| replace PointCutMix-K by AMS on CCN (ablation) | CCN · ModelNet40 white-box | LAUA: 68.30 -> 77.72 | Improved |
| replace AMS with AT-SMA on CCN (ablation) | CCN · ModelNet40 white-box | SMA-40 accuracy: 77.72 -> 84.17 | Hurt |
| add AMS training to DGCNN instead of AT-BIM | DGCNN · ModelNet40 white-box | AAUA: 78.76 -> 83.91 | 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.
- DUP-Net — DUP-Net: Denoiser and Upsampler Network for 3D Adversarial Point Clouds Defense (ICCV 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
- Benchmarking and Analyzing Robust Point Cloud Recognition: Bag of Tricks for Defending Adversarial Examples (ICCV 2023 · similarity 0.94)
- CAP: Robust Point Cloud Classification via Semantic and Structural Modeling (CVPR 2023 · similarity 0.93)
- Minimal Adversarial Examples for Deep Learning on 3D Point Clouds (ICCV 2021 · similarity 0.91)
- A Critical Revisit of Adversarial Robustness in 3D Point Cloud Recognition with Diffusion-Driven Purification (ICML 2023 · similarity 0.91)
- Robust Adversarial Objects against Deep Learning Models (AAAI 2020 · similarity 0.91)
- Curvature-Invariant Adversarial Attacks for 3D Point Clouds (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.
