Distracting Downpour: Adversarial Weather Attacks for Motion Estimation
J. Schmalfuss, L. Mehl, and A. Bruhn. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), page 10106-10116. (October 2023)
Abstract
Current adversarial attacks on motion estimation, or optical flow, optimize small per-pixel perturbations, which are unlikely to appear in the real world. In contrast, adverse weather conditions constitute a much more realistic threat scenario. Hence, in this work, we present a novel attack on motion estimation that exploits adversarially optimized particles to mimic weather effects like snowflakes, rain streaks or fog clouds. At the core of our attack framework is a differentiable particle rendering system that integrates particles (i) consistently over multiple time steps (ii) into the 3D space (iii) with a photo-realistic appearance. Through optimization, we obtain adversarial weather that significantly impacts the motion estimation. Surprisingly, methods that previously showed good robustness towards small per-pixel perturbations are particularly vulnerable to adversarial weather. At the same time, augmenting the training with non-optimized weather increases a method's robustness towards weather effects and improves generalizability at almost no additional cost. Our code is available at https://github.com/cv-stuttgart/DistractingDownpour.
%0 Conference Paper
%1 Schmalfuss_2023_ICCV
%A Schmalfuss, Jenny
%A Mehl, Lukas
%A Bruhn, Andrés
%B Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)
%D 2023
%K
%P 10106-10116
%T Distracting Downpour: Adversarial Weather Attacks for Motion Estimation
%U https://openaccess.thecvf.com/content/ICCV2023/html/Schmalfuss_Distracting_Downpour_Adversarial_Weather_Attacks_for_Motion_Estimation_ICCV_2023_paper.html
%X Current adversarial attacks on motion estimation, or optical flow, optimize small per-pixel perturbations, which are unlikely to appear in the real world. In contrast, adverse weather conditions constitute a much more realistic threat scenario. Hence, in this work, we present a novel attack on motion estimation that exploits adversarially optimized particles to mimic weather effects like snowflakes, rain streaks or fog clouds. At the core of our attack framework is a differentiable particle rendering system that integrates particles (i) consistently over multiple time steps (ii) into the 3D space (iii) with a photo-realistic appearance. Through optimization, we obtain adversarial weather that significantly impacts the motion estimation. Surprisingly, methods that previously showed good robustness towards small per-pixel perturbations are particularly vulnerable to adversarial weather. At the same time, augmenting the training with non-optimized weather increases a method's robustness towards weather effects and improves generalizability at almost no additional cost. Our code is available at https://github.com/cv-stuttgart/DistractingDownpour.
@inproceedings{Schmalfuss_2023_ICCV,
abstract = {Current adversarial attacks on motion estimation, or optical flow, optimize small per-pixel perturbations, which are unlikely to appear in the real world. In contrast, adverse weather conditions constitute a much more realistic threat scenario. Hence, in this work, we present a novel attack on motion estimation that exploits adversarially optimized particles to mimic weather effects like snowflakes, rain streaks or fog clouds. At the core of our attack framework is a differentiable particle rendering system that integrates particles (i) consistently over multiple time steps (ii) into the 3D space (iii) with a photo-realistic appearance. Through optimization, we obtain adversarial weather that significantly impacts the motion estimation. Surprisingly, methods that previously showed good robustness towards small per-pixel perturbations are particularly vulnerable to adversarial weather. At the same time, augmenting the training with non-optimized weather increases a method's robustness towards weather effects and improves generalizability at almost no additional cost. Our code is available at https://github.com/cv-stuttgart/DistractingDownpour.},
added-at = {2024-03-15T12:03:53.000+0100},
author = {Schmalfuss, Jenny and Mehl, Lukas and Bruhn, Andr\'es},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/23e2073a3d32f829425fa4538436fdae1/visus},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
interhash = {66326570f27a01e95abd0a337c4325e6},
intrahash = {3e2073a3d32f829425fa4538436fdae1},
keywords = {},
month = {October},
pages = {10106-10116},
timestamp = {2024-03-15T12:03:53.000+0100},
title = {Distracting Downpour: Adversarial Weather Attacks for Motion Estimation},
url = {https://openaccess.thecvf.com/content/ICCV2023/html/Schmalfuss_Distracting_Downpour_Adversarial_Weather_Attacks_for_Motion_Estimation_ICCV_2023_paper.html},
year = 2023
}