A Perturbation Constrained Adversarial Attack for Evaluating the Robustness of Optical Flow

Recent optical flow methods are almost exclusively judged in terms of accuracy, while analyzing their robustness is often neglected. Although adversarial attacks offer a useful tool to perform such an analysis, current attacks on optical flow methods rather focus on real-world attacking scenarios than on a worst case robustness assessment. Hence, in this work, we propose a novel adversarial attack - the Perturbation Constrained Flow Attack (PCFA) - that emphasizes destructivity over applicability as a real-world attack. More precisely, PCFA is a global attack that optimizes adversarial perturbations to shift the predicted flow towards a specified target flow, while keeping the L2 norm of the perturbation below a chosen bound. Our experiments not only demonstrate PCFA's applicability in white- and black-box settings, but also show that it finds stronger adversarial samples for optical flow than previous attacking frameworks. Moreover, based on these strong samples, we provide the first common ranking of optical flow methods in the literature considering both prediction quality and adversarial robustness, indicating that high quality methods are not necessarily robust. Our source code will be publicly available.