Structured Adversarial Camouflage via Voronoi Diagrams
Researchers have developed a new method for creating adversarial camouflage patterns using Voronoi diagrams, which optimizes seed-point locations for printable, structured patterns. This technique aims to be more visually plausible and computationally efficient than pixel-wise adversarial patches. When tested on person detection using the COCO dataset, the camouflage significantly degraded detector performance, even transferring across different detector families like YOLOv9 through YOLOv12. AI
IMPACT This research demonstrates a novel approach to adversarial attacks that could impact the robustness of computer vision systems.
- Hugging Face
- YOLOv10
- DagsHub
- YOLOv11
- COCO
- alphaXiv
- Connected Papers
- Litmaps
- scite Smart Citations
- YOLOv12
- Gotit.pub
- Institut National de Recherche en Informatique et en Automatique
- YOLOv9
- CatalyzeX Code Finder for Papers
- International Conference on Military Communication and Information Systems (ICMCIS)
- information systems technology
- Voronoi diagram
- 3DPeople