Research map / Point clouds / Segmentation and detection
Explainability-aware Frustum Attack: Exposing Structural Vulnerabilities in LiDAR-Based 3D Object Detectors
Chengzeng You, Binbin Xu, Soteris Demetriou
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 |
|---|---|---|---|
| select frustums by SALL saliency (Critical Frustum First) instead of randomly (ablation) | EFA, KITTI Car PointPillars · KITTI Car recall | recall: recall 0.0 at 30 frustums vs 50 frustums | Improved |
| use Shift instead of Remove perturbation (ablation) | EFA with Critical Frustum First · KITTI Car | ASR (%) | No clear effect |
| guide frustum choice with SALL saliency instead of HFR | HFR · nuScenes Car (20 frustums) | ASR (%): 78.29 -> 96.55 | Improved |
| guide frustum choice with SALL saliency instead of PRA | PRA · nuScenes Car (20 frustums) | ASR (%): 73.47 -> 96.55 | Improved |
| guide frustum choice with SALL saliency instead of HFR | HFR · KITTI Pedestrian (3 frustums) | ASR (%): 74.25 -> 97.00 | Improved |
| guide A-HFR with SALL critical regions | A-HFR · KITTI Pedestrian at 50 km/h | ASR (%): 67.38 -> 92.70 | Improved |
| use PointPillars map instead of native SECOND map (ablation) | SECOND attack · KITTI Car (30 frustums) | ASR (%): 88.12 -> 94.84 | Improved |
| use joint detector saliency map instead of SECOND-only map (ablation) | SECOND attack · KITTI Car (30 frustums) | ASR (%): 88.12 -> 95.29 | Improved |
| use joint detector map instead of SECOND-only map (ablation) | PointPillars attack · KITTI Car (30 frustums) | ASR (%): 97.63 -> 99.44 | Improved |
| use joint dataset map instead of KITTI-only map (ablation) | nuScenes attack · nuScenes Car (20 frustums) | ASR (%): 96.64 -> 96.90 | Improved |
| use joint dataset map instead of KITTI-only map (ablation) | KITTI attack · KITTI Car (20 frustums) | ASR (%): 88.42 -> 88.15 | No clear effect |
| use Zero instead of Nearest Corner IG base point (ablation) | SALL · KITTI Car (20 frustums) | ASR (%): 87.73 -> 90.66 | No clear effect |
Closest papers by meaning
- FFAM: Feature Factorization Activation Map for Explanation of 3D Detectors (NeurIPS 2024 · similarity 0.88)
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds (CVPR 2024 · similarity 0.88)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.87)
- A New Adversarial Perspective for LiDAR-based 3D Object Detection (AAAI 2025 · similarity 0.86)
- Fusion Is Not Enough: Single Modal Attacks on Fusion Models for 3D Object Detection (ICLR 2024 · similarity 0.86)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · 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.
