Towards Adversarially Robust Object Detection
This work addresses the vulnerability of object detection systems to adversarial attacks, which is critical for practical applications, though it is an incremental improvement in robustness.
The paper tackled the problem of adversarial robustness in object detection models by developing an adversarial training approach that leverages multiple attack sources, achieving verified effectiveness on PASCAL-VOC and MS-COCO datasets.
Object detection is an important vision task and has emerged as an indispensable component in many vision system, rendering its robustness as an increasingly important performance factor for practical applications. While object detection models have been demonstrated to be vulnerable against adversarial attacks by many recent works, very few efforts have been devoted to improving their robustness. In this work, we take an initial attempt towards this direction. We first revisit and systematically analyze object detectors and many recently developed attacks from the perspective of model robustness. We then present a multi-task learning perspective of object detection and identify an asymmetric role of task losses. We further develop an adversarial training approach which can leverage the multiple sources of attacks for improving the robustness of detection models. Extensive experiments on PASCAL-VOC and MS-COCO verified the effectiveness of the proposed approach.