Perspective-Invariant Attack With Enhanced Transferability of Adversarial Examples
This work proposes a Perspective-Invariant Attack (PIA), which introduces a multi-DOF vertex sampling strategy that systematically covers the perspective transformation hierarchy from 2-DOF translation to 8-DOF projective mapping, and proposes PIA-Mix, a generic extension that maintains a complementary transformation pool and efficiently combines the authors' perspective transformation with auxiliary methods for improved transferability.