Research map / Adversarial robustness and security / Attacks on text and graphs
Physically Realizable Adversarial Examples for LiDAR Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train PIXOR with random rooftop-mesh data augmentation | PIXOR · KITTI validation, arbitrary rooftop object | attack success rate: 77.3% -> 14.4% | Helped |
| Train PIXOR with adversarial training | PIXOR · KITTI validation, arbitrary rooftop object | attack success rate: 77.3% -> 10.8% | Helped |
| Train PIXOR with random rooftop-mesh data augmentation | PIXOR · KITTI validation | AP: 74.37 -> 74.92 | Helped |
| Train PIXOR with adversarial training | PIXOR · KITTI validation | AP: 74.37 -> 73.97 | Hurt |
| Use a universal learned rooftop mesh rather than an initial icosphere (ablation) | PIXOR (density) · KITTI validation | recall | Helped |
| Use black-box genetic optimization instead of white-box gradient optimization (ablation) | PIXOR (density) · KITTI validation | recall | No clear effect |
| Transfer a PIXOR-1 attack to a different-seed PIXOR-2 | PIXOR · PIXOR-1 → PIXOR-2 | attack success rate: 77.3% -> 53.9% | Mixed |
| Use PointRCNN as the target detector rather than PIXOR | Universal rooftop adversarial mesh · KITTI validation | attack success rate: 77.3% -> 32.3% | Hurt |
| Use PointPillar as the target detector rather than PIXOR | Universal rooftop adversarial mesh · KITTI validation | attack success rate: 77.3% -> 57.5% | Hurt |
| Use a point-level representation and reasoning scheme rather than PIXOR occupancy voxels | LiDAR detector · KITTI validation | attack success rate: 77.3% -> 32.3% | Mixed |
| Constrain the mesh to resemble common rooftop objects (ablation) | PIXOR · KITTI validation, common-object rooftop attacks | attack success rate: 23.3%; 68.6%; 59.5%; 48.0%; 54.1%; 32.4% | Mixed |
Closest papers by meaning
- A New Adversarial Perspective for LiDAR-based 3D Object Detection (AAAI 2025 · similarity 0.91)
- Seeing is Not Believing: Adversarial Natural Object Optimization for Hard-Label 3D Scene Attacks (CVPR 2025 · similarity 0.87)
- Enduring, Efficient and Robust Trajectory Prediction Attack in Autonomous Driving via Optimization-Driven Multi-Frame Perturbation Framework (CVPR 2025 · similarity 0.87)
- Fusion Is Not Enough: Single Modal Attacks on Fusion Models for 3D Object Detection (ICLR 2024 · similarity 0.86)
- Physical 3D Adversarial Attacks against Monocular Depth Estimation in Autonomous Driving (CVPR 2024 · similarity 0.86)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (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.