Researchers have developed AdvTiles, a novel framework for creating physical adversarial camouflage clothing designed to evade person detection systems. This method utilizes learnable tiles and a Straight-through Gumbel-Softmax estimator for differentiable tile selection, allowing for fine-grained control over adversarial patterns and their arrangement. The system also incorporates 3D Gaussian Splatting for robust optimization across various viewpoints, lighting, and backgrounds. Experiments show AdvTiles achieves an 86.2% average attack success rate, outperforming existing methods, and has been validated in real-world scenarios with wearable clothing. AI
IMPACT This research could lead to new methods for evading AI-powered surveillance and security systems.
RANK_REASON This is a research paper detailing a new method for adversarial attacks in computer vision. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
