Fooling Thermal Infrared Pedestrian Detectors in Real World Using Small Bulbs

Thermal infrared detection systems play an important role in many areas such as night security, autonomous driving, and body temperature detection. They have the unique advantages of passive imaging, sensitivity penetration. But security these themselves has not been fully explored, which poses risks applying systems. We propose a physical attack method with small bulbs on board against state of-the-art pedestrian detectors. Our goal is to make detectors unable detect real-world pedestrians. Towards this goal, we ?rst showed that it possible use two kinds patches detector based YOLOv3. The average precision (AP) dropped by 64.12% digital world, while blank same size caused AP drop 29.69% only. After that, designed manufactured successfully attacked YOLOv3 real world. In recorded videos, target 34.48%, 14.91% With ensemble techniques, had good transferability unseen