Research map / Adversarial robustness and security / Backdoor and poisoning
6DAttack: Backdoor Attacks in the 6DoF Pose Estimation
Jihui Guo, Zongmin Zhang, Zhen Sun, Yuhao Yang, Jinlin Wu, Fu Zhang, Xinlei He
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 |
|---|---|---|---|
| embed 3D object triggers with edited pose labels (5-20% poisoning) | DenseFusion · Cracker Box with Mouse trigger | ASR: 0% -> 100% | Improved |
| increase poison ratio from 5% to 20% (ablation) | DenseFusion · Cracker Box clean ADD-C | ADD-C: 100.00% -> 97.56% | Hurt |
| increase poison ratio from 5% to 15% (ablation) | DenseFusion · Cracker Box triggered ADD-P | ADD-P: 56.23% -> 97.70% | Improved |
| increase poison ratio from 5% to 20% (ablation) | PVNet · Can with Mouse trigger clean ADD-C | ADD-C: 53.97% -> 28.87% | Hurt |
| increase poison ratio from 5% to 20% (ablation) | PVNet · Can with Mouse trigger triggered ADD-P | ADD-P: 8.37% -> 18.01% | Improved |
| increase poison ratio from 5% to 20% (ablation) | PoseDiffusion · CO3D apple clean ADD-C | ADD-C: 71.29% -> 71.27% | No clear effect |
| fine-tune poisoned model with extra clean data (defense) | PVNet LINEMOD 5% poisoned · triggered scenes | ASR | No clear effect |
| fine-tune poisoned model with extra clean data (defense) | PVNet LINEMOD 5% poisoned · triggered scenes | ADD-P | Mixed |
Closest papers by meaning
- Finding Naturally Occurring Physical Backdoors in Image Datasets (NeurIPS 2022 · similarity 0.85)
- Detecting Backdoors in Object Detection via Pre-NMS Prediction Distribution Shift (ECCV 2026 · similarity 0.84)
- Few-Shot Backdoor Attacks on Visual Object Tracking (ICLR 2022 · similarity 0.84)
- Single-Stage 6D Object Pose Estimation (CVPR 2020 · similarity 0.84)
- Not All Prompts Are Secure: A Switchable Backdoor Attack Against Pre-trained Vision Transfomers (CVPR 2024 · similarity 0.83)
- BadVLA: Towards Backdoor Attacks on Vision-Language-Action Models via Objective-Decoupled Optimization (NeurIPS 2025 · similarity 0.83)
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.
