Research map / Adversarial robustness and security / Attacks on text and graphs
Design and Interpretation of Universal Adversarial Patches in Face Detection
Xiao Yang, Fangyun Wei, Hongyang Zhang, Jun Zhu
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 |
|---|---|---|---|
| optimize noise patch with Patch-IoU instead of pasting it unoptimized | Noise patch · WIDER FACE val All | recall (%): 14.2 -> 1.4 | Improved |
| optimize face patch FaceC with Patch-IoU | FaceC patch · WIDER FACE val All | recall (%): 10.2 -> 0.0 | Improved |
| optimize noise patch with Patch-IoU | Noise patch · WIDER FACE val All | precision (%): 99.4 -> 7.3 | Improved |
| use Patch-Score instead of Patch-IoU (add score-based sample selection) | Patch-IoU · WIDER FACE val All | precision (%): 7.3 -> 99.0 | Improved |
| add focal modulating factor to Patch-Score (ablation) | Patch-Score · WIDER FACE val All | recall (%): 6.7 -> 6.1 | Improved |
| add IoU-based combination selection to Patch-Score (ablation) | Patch-Score · WIDER FACE val All | recall (%): 6.7 -> 5.4 | Improved |
| use only half of the Patch-IoU patch (top) (ablation) | Patch-IoU · WIDER FACE val All | recall (%): 1.4 -> 12.5 | Hurt |
| use only two thirds of the patch (remove right third) (ablation) | Patch-IoU · WIDER FACE val All | recall (%): 1.4 -> 4.6 | Hurt |
| shrink Patch-Combination patch from 90% to 60% (ablation) | Patch-Combination · WIDER FACE val All | recall (%): 9.2 -> 17.2 | Hurt |
| transfer patch trained with Faster R-CNN to SLN instead of SSD (ablation) | SLN detector · WIDER FACE val All | recall (%): 8.6 -> 1.2 | Improved |
| place Patch-Combination at middle instead of top (ablation) | Patch-Combination · WIDER FACE val All | recall (%): 5.3 -> 4.0 | Improved |
| transfer Patch-Combination to ResNet-50 black-box | ResNet-50 detector · WIDER FACE val All | recall (%): 32.5 -> 12.8 | Improved |
Closest papers by meaning
- Revisiting Adversarial Patch Defenses on Object Detectors: Unified Evaluation, Large-Scale Dataset, and New Insights (ICCV 2025 · similarity 0.89)
- Segment and Complete: Defending Object Detectors Against Adversarial Patch Attacks With Robust Patch Detection (CVPR 2022 · similarity 0.88)
- A Unified, Resilient, and Explainable Adversarial Patch Detector (CVPR 2025 · similarity 0.88)
- Revisiting Adversarial Patches for Designing Camera-Agnostic Attacks against Person Detection (NeurIPS 2024 · similarity 0.87)
- PAD: Patch-Agnostic Defense against Adversarial Patch Attacks (CVPR 2024 · similarity 0.87)
- DAP: A Dynamic Adversarial Patch for Evading Person Detectors (CVPR 2024 · similarity 0.86)
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.
