Get Started
Home
Topics
Search
Library
Research questionHow can adversarial perturbations transfer to unseen semantic-segmentation models while accounting for dense spatial and class-wise structure?A perturbation crafted on a surrogate segmentation model must mislead an unseen target model. This is difficult because dense prediction depends on spatially organized and class-specific representations, not only output scores.
AI
Alignment & Safety
Computer Vision
Image & Video Processing
Machine Learning
Research Paper
Latest papersRecent research connected to this question, newest first.FSPGD: Rethinking Black-box Attacks on Semantic SegmentationThe paper proposes a feature-space transfer attack based on discrepancies in intermediate representations and evaluates it across CNN and Transformer segmentation backbones on Pascal VOC 2012 and Cityscapes. It also examines robustness to unseen attacks after adversarial training; evidence is limited to those datasets, architectures, and attack settings.research paper · Sep 4, 2026
Related questions
Can small targeted grayscale patches force chosen semantics in infrared vision-language models across tasks?How can histopathology models withstand adversarial perturbations without erasing diagnostic tissue structures?How can black-box adversarial attacks disrupt image-based continuous-control agents with fewer environment queries?Can black-box attackers identify and reconstruct prompts supposedly removed from language models without knowing them in advance?