AI RESEARCH
TsallisPGD: Adaptive Gradient Weighting for Adversarial Attacks on Semantic Segmentation
arXiv CS.LG
•
ArXi:2605.03405v1 Announce Type: cross Attacking semantic segmentation models is significantly harder than image classification models because an attacker must flip thousands of pixel predictions simultaneously. Standard pixel-wise cross-entropy (CE) is ill-suited to this setting: it tends to overemphasize already-misclassified pixels, which slows optimization and overstates model robustness. To address these issues, we