Research map / Adversarial robustness and security / Attacks on text and graphs
Adversarial Texture for Fooling Person Detectors in the Physical World
Zhanhao Hu, Siyuan Huang, Xiaopei Zhu, Fuchun Sun, Bo Zhang, 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 |
|---|---|---|---|
| Use TC-EGA texture instead of random texture | Physical person-detector attacks · YOLOv2 | mASR: 0.087 -> 0.743 | Improved |
| Use TC-EGA texture instead of random texture | Physical person-detector attacks · YOLOv3 | mASR: 0.000 -> 0.701 | Improved |
| Use TC-EGA texture instead of random texture | Physical person-detector attacks · Faster R-CNN | mASR: 0.000 -> 0.930 | Improved |
| Use TC-EGA texture instead of random texture | Physical person-detector attacks · Mask R-CNN | mASR: 0.000 -> 0.855 | Improved |
| Optimize the latent variable after training the expandable generator (ablation) | TC-EGA · Inria test set (YOLOv2, digital patch attack) | AP: 0.470 -> 0.362 | Improved |
| Use a trained FCN generator with toroidal cropping rather than directly optimizing the texture (ablation) | TC-EGA · Inria test set (YOLOv2, digital patch attack) | AP: 0.664 -> 0.362 | Improved |
| Use toroidal cropping instead of random cropping of a fixed-size large patch (ablation) | Texture optimization · Inria test set (YOLOv2, digital patch attack) | AP: 0.606 -> 0.362 | Improved |
| Reduce the fixed optimized patch size from 900×900 to 300×300 (ablation) | Random Cropping Attack · Inria test set (YOLOv2, digital patch attack) | AP: 0.855 -> 0.606 | Improved |
| Tile AdvPatch into a repeated texture | AdvPatch · Inria test set (YOLOv2, digital patch attack) | AP: 0.352 -> 0.827 | Hurt |
| Tile AdvTshirt into a repeated texture | AdvTshirt · Inria test set (YOLOv2, digital patch attack) | AP: 0.744 -> 0.844 | Hurt |
| Apply the texture to a larger dress rather than a T-shirt | TC-EGA clothing · Physical YOLOv2 test | mASR: 0.771 -> 0.893 | Improved |
| Increase camera distance from 1.6 m to 7.0 m | TC-EGA T-shirt · Physical YOLOv2 test | mASR: 0.791 -> 0.257 | Hurt |
Closest papers by meaning
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.87)
- DAP: A Dynamic Adversarial Patch for Evading Person Detectors (CVPR 2024 · similarity 0.86)
- UV-Attack: Physical-World Adversarial Attacks on Person Detection via Dynamic-NeRF-based UV Mapping (ICLR 2025 · similarity 0.85)
- Design2Cloth: 3D Cloth Generation from 2D Masks (CVPR 2024 · similarity 0.84)
- Adversarial Perturbations Cannot Reliably Protect Artists From Generative AI (ICLR 2025 · similarity 0.83)
- Towards Squeezing-Averse Virtual Try-On via Sequential Deformation (AAAI 2024 · 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.
