PB-UAP is a universal adversarial perturbation for semantic segmentation that combines feature deviation with low-frequency scattering and reduces segmentation mIoU to between 3 and 19 percent on tested models.
Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
PB-UAP: Hybrid Universal Adversarial Attack For Image Segmentation
PB-UAP is a universal adversarial perturbation for semantic segmentation that combines feature deviation with low-frequency scattering and reduces segmentation mIoU to between 3 and 19 percent on tested models.