Research map / Adversarial robustness and security / Attacks on text and graphs
Making an Invisibility Cloak: Real World Adversarial Attacks on Object Detectors
Zuxuan Wu, Ser-Nam Lim, Larry S. Davis, Tom Goldstein
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 |
|---|---|---|---|
| add adversarially learned YOLOv2 patch instead of no patch | YOLOv2 detector · COCO person, digital patch attack | AP (%): 63.6 -> 10.7 | Improved |
| use learned YOLOv2 patch instead of Seurat patch | YOLOv2 detector · COCO person, digital patch attack | AP (%): 43.4 -> 10.7 | Improved |
| use learned YOLOv2 patch instead of Grey patch | YOLOv2 detector · COCO person, digital patch attack | AP (%): 48 -> 10.7 | Improved |
| train patch on ensemble Ens2 instead of R50-FPN alone, victim YOLOv2 | R50-FPN patch · COCO person, digital patch attack | AP (%): 22.6 -> 17.5 | Mixed |
| transfer YOLOv2 patch to R50-FPN victim | YOLOv2 patch · COCO person, digital patch attack | AP (%): 82.2 -> 21.5 | Improved |
| transfer R50-C4 patch to YOLOv2 victim | R50-C4 patch · COCO person, digital patch attack | AP (%): 63.6 -> 37.9 | No clear effect |
| attack retrained victim YOLOv2-r instead of pretrained with YOLOv2 patch | YOLOv2 patch · COCO person, digital patch attack | AP (%): 10.7 -> 7.5 | Improved |
| train person patch, test on person class in VOC | Person patch · VOC person class | AP (%): 83.3 -> 5.5 | Improved |
| train patch on person instead of grey patch, test across VOC classes | Person patch · VOC classes | AP (%): 35.8 -> 5.5 | Improved |
| add TPS transformations during patch training (ablation) | YOLOv2 patch · Physical posters and shirts | AP and success rate | Hurt |
| add augmentations during patch training (ablation) | YOLOv2 patch · Physical posters | AP | Improved |
| train patch on Inria instead of COCO | YOLOv2 patch · Physical posters | success rate | Mixed |
Closest papers by meaning
- Transferable Adversarial Attacks for Object Detection Using Object-Aware Significant Feature Distortion (AAAI 2024 · similarity 0.91)
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.88)
- GLOW: Global Layout Aware Attacks on Object Detection (CVPR 2024 · similarity 0.88)
- Full-Distance Evasion of Pedestrian Detectors in the Physical World (NeurIPS 2024 · similarity 0.87)
- Revisiting Adversarial Patches for Designing Camera-Agnostic Attacks against Person Detection (NeurIPS 2024 · similarity 0.87)
- Can't Slow Me Down: Learning Robust and Hardware-Adaptive Object Detectors against Latency Attacks for Edge Devices (CVPR 2025 · 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.
