Using Frequency Attention to Make Adversarial Patch Powerful Against Person Detector

Deep neural networks (DNNs) are vulnerable to adversarial attacks. In particular, object detectors may be attacked by applying a particular patch the image. However, because will lose information during shrinking process, most existing approaches that employ patches attack would diminish success rate on small and medium targets. This paper proposes Frequency Module(FRAN), frequency-domain attention module for guiding generation solve this problem. We use FRAN allow patched rely low-frequency signals which lost less in process. is first study introduce frequency domain optimize capabilities of patches. Our method increases rates targets 4.18% 3.89%, respectively, over state-of-the-art fooling human detector while assaulting YOLOv3 without reducing big