1 paper · 1 filter
Di Yang, Yihao Huang, Qing Guo +4
The adversarial patch attack aims to fool image classifiers within a bounded, contiguous region of arbitrary changes, posing a real threat to computer vision systems (e.g., autonom…