Research map / Adversarial robustness and security / Attacks on text and graphs
Infrared Invisible Clothing: Hiding From Infrared Detectors at Multiple Angles in Real World
Xiaopei Zhu, Zhanhao Hu, Siyuan Huang, Jianmin Li, Xiaolin Hu
Check 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 |
|---|---|---|---|
| Optimize a tiled binary QR pattern with random cropping, cloth deformation, and EOT | Infrared clothing pattern for YOLOv3 · PEOPLE_FLIR test (digital) | AP decrease: 87.7% | Improved |
| Replace optimized pattern with a random QR code pattern (ablation) | Infrared clothing pattern for YOLOv3 · PEOPLE_FLIR test (digital) | AP decrease: 57.9% | Hurt |
| Replace optimized pattern with a blank pattern (ablation) | Infrared clothing pattern for YOLOv3 · PEOPLE_FLIR test (digital) | AP decrease: 30.1% | Hurt |
| Use optimized QR pattern clothing rather than control clothing | Physical infrared clothing · Physical-world videos | AP decrease: 64.6% | Improved |
| Replace the optimized physical pattern with random QR clothing, fully heat-insulated clothing, or ordinary clothing (ablation) | Physical infrared clothing · Physical-world videos | AP decrease: 28.3%; 22.8%; 4.0% | Hurt |
| Optimize against an ensemble of detectors rather than YOLOv3 alone | Pattern transfer to unseen detectors · PEOPLE_FLIR test (digital), Deformable DETR, RetinaNet, and Libra-RCNN | AP decrease: 13.7% -> 24.7%; 25.3% -> 58.2%; 32.7% -> 66.9% | Improved |
| Use a 10×10 basic patch instead of 20×20 (ablation) | YOLOv3 infrared pattern optimization · PEOPLE_FLIR test (digital) | AP decrease: 87.7% -> 64.9% | Hurt |
| Vary basic patch size across 30×30, 40×40, and 50×50 rather than 20×20 (ablation) | YOLOv3 infrared pattern optimization · PEOPLE_FLIR test (digital) | AP decrease: 87.7% -> 85.6%; 83.6%; 83.4% | Hurt |
| Increase λ from 0 to 1 to penalize black pixels (ablation) | YOLOv3 infrared pattern optimization · PEOPLE_FLIR test (digital) | AP decrease: 89.2% -> 88.4% -> 88.2% -> 87.7% -> 87.6% -> 87.2% | Mixed |
| Increase λ from 0 to 1 to penalize black pixels (ablation) | YOLOv3 infrared pattern optimization · PEOPLE_FLIR test (digital) | Black pixel ratio: 52.0% -> 49.8% -> 48.2% -> 47.3% -> 45.3% -> 44.3% | Improved |
| Apply the tested defense methods | Digital infrared clothing attack · PEOPLE_FLIR test (digital) | AP: 12.3% to 36.8%; attack still lowered AP by 63.2% | Mixed |
Closest papers by meaning
- Physical Adversarial Examples for Person Detectors in Thermal Images Based on 3D Modeling (TPAMI 2025 · similarity 0.94)
- Infrared Adversarial Car Stickers (CVPR 2024 · similarity 0.84)
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.84)
- AdvDisplay: Adversarial Display Assembled by Thermoelectric Cooler for Fooling Thermal Infrared Detectors (AAAI 2025 · similarity 0.82)
- Design2Cloth: 3D Cloth Generation from 2D Masks (CVPR 2024 · similarity 0.82)
- DAP: A Dynamic Adversarial Patch for Evading Person Detectors (CVPR 2024 · similarity 0.81)
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.
