Research map / Adversarial robustness and security / Attacks on text and graphs
ACTIVE: Towards Highly Transferable 3D Physical Camouflage for Universal and Robust Vehicle Evasion
Naufal Suryanto, Yongsu Kim, Harashta Tatimma Larasati, Hyoeun Kang, Thi-Thu-Huong Le, Yoonyoung Hong, Hunmin Yang, Se-Yoon Oh, Howon Kim
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 |
|---|---|---|---|
| use triplanar mapping with neural texture renderer instead of repeated texture projection (DTA) | ACTIVE · UE4 physical simulation YOLOv3 | car AP@0.5 (%): 41.95 -> 7.29 | Improved |
| add triplanar mapping module (ablation) | ACTIVE with stealth loss · YOLOv3 | car AP@0.5 (%): 60.13 -> 43.70 | Improved |
| add random output augmentation (ROA) (ablation) | ACTIVE with stealth loss · YOLOv3 | car AP@0.5 (%): 60.13 -> 38.85 | Improved |
| use stealth loss instead of DTA attack loss (ablation) | ACTIVE full modules · YOLOv3 | car AP@0.5 (%): 24.66 -> 22.48 | Improved |
| add smooth loss to stealth loss (ablation) | ACTIVE · YOLOv3 | car AP@0.5 (%): 22.48 -> 20.21 | Improved |
| add camouflage loss to stealth loss (ablation) | ACTIVE · YOLOv3 | car AP@0.5 (%): 22.48 -> 24.34 | Hurt |
| add both smooth and camouflage losses (ablation) | ACTIVE · YOLOv3 | car AP@0.5 (%): 22.48 -> 28.22 | Hurt |
| transfer texture to transformer detectors and YOLOv7 (black-box) | ACTIVE · Deformable DETR, YOLOv7 | car AP@0.5 (%): 83.83 -> 14.69 (Deformable DETR) | Improved |
| apply texture to scaled real-world car | ACTIVE · Real-world YOLOv7 | car AP@0.5 (%): 95.00 -> 48.75 | Improved |
Closest papers by meaning
- CNCA: Toward Customizable and Natural Generation of Adversarial Camouflage for Vehicle Detectors (NeurIPS 2024 · similarity 0.91)
- RAUCA: A Novel Physical Adversarial Attack on Vehicle Detectors via Robust and Accurate Camouflage Generation (ICML 2024 · similarity 0.91)
- PhyCamo: A Robust Physical Camouflage via Contrastive Learning for Multi-View Physical Adversarial Attack (AAAI 2025 · similarity 0.88)
- Camouflage Anything: Learning to Hide using Controlled Out-painting and Representation Engineering (CVPR 2025 · similarity 0.86)
- Physical 3D Adversarial Attacks against Monocular Depth Estimation in Autonomous Driving (CVPR 2024 · similarity 0.85)
- Infrared Adversarial Car Stickers (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.
