Improving the Imperceptibility of Adversarial Examples Based on Weakly Perceptual Perturbation in Key Regions

<div>The framework of our WPAE consists of a generator, a discriminator, a filter, and a positioning module (<i>M</i>), attacking the target faster RCNN model. The pretrained convolutional network is used to extract the content representation from the image.</div>

Security and Communication Networks

fig4

Figure 4

Figure 4: Improving the Imperceptibility of Adversarial Examples Based on Weakly Perceptual Perturbation in Key Regions 