Sep, 2023

抵御对抗性补丁攻击的 RGB-D 物体识别系统强化

TL;DRRGB-D object recognition systems are vulnerable to adversarial examples, and color features contribute to this weakness, making the network more sensitive to perturbations. To address this issue, a defense mechanism is proposed, which improves the performance of RGB-D systems against adversarial examples and exceeds the effectiveness of adversarial training.