FCA: Learning a 3D Full-Coverage Vehicle Camouflage for Multi-View Physical Adversarial Attack

Physical adversarial attacks in object detection have attracted increasing attention. However, most previous works focus on hiding the objects from detector by generating an individual patch, which only covers planar part of vehicle’s surface and fails to attack physical scenarios for multi-view, long-distance partially occluded objects. To bridge gap between digital attacks, we exploit full 3D vehicle propose a robust Full-coverage Camouflage Attack (FCA) fool detectors. Specifically, first try rendering nonplanar camouflage texture over surface. mimic real-world environment conditions, then introduce transformation function transfer rendered camouflaged into photo-realistic scenario. Finally, design efficient loss optimize texture. Experiments show that full-coverage can not outperform state-of-the-art methods under various test cases but also generalize different environments, vehicles,