The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
L'estimation de la profondeur monoculaire s'est considérablement améliorée grâce au développement des réseaux de neurones profonds (DNN). Cependant, des études récentes ont révélé que les DNN pour l’estimation monoculaire de la profondeur contiennent des vulnérabilités qui peuvent conduire à une mauvaise estimation lorsque des perturbations sont ajoutées aux entrées. Cette étude examine si les DNN pour l'estimation de la profondeur monoculaire sont vulnérables à une mauvaise estimation lorsqu'une lumière structurée est projetée sur un objet à l'aide d'un vidéoprojecteur. À cette fin, cette étude propose une méthode d’attaque contradictoire évolutive avec un schéma d’évaluation multi-fidélité qui permet de créer des exemples contradictoires dans des conditions de boîte noire tout en supprimant le coût de calcul. Des expériences dans des scènes simulées et réelles ont montré que le modèle de lumière conçu a amené un DNN à mal estimer les objets comme s'ils s'étaient déplacés vers l'arrière.
Renya DAIMO
Kagoshima University
Satoshi ONO
Kagoshima University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Renya DAIMO, Satoshi ONO, "Projection-Based Physical Adversarial Attack for Monocular Depth Estimation" in IEICE TRANSACTIONS on Information,
vol. E106-D, no. 1, pp. 31-35, January 2023, doi: 10.1587/transinf.2022MUL0001.
Abstract: Monocular depth estimation has improved drastically due to the development of deep neural networks (DNNs). However, recent studies have revealed that DNNs for monocular depth estimation contain vulnerabilities that can lead to misestimation when perturbations are added to input. This study investigates whether DNNs for monocular depth estimation is vulnerable to misestimation when patterned light is projected on an object using a video projector. To this end, this study proposes an evolutionary adversarial attack method with multi-fidelity evaluation scheme that allows creating adversarial examples under black-box condition while suppressing the computational cost. Experiments in both simulated and real scenes showed that the designed light pattern caused a DNN to misestimate objects as if they have moved to the back.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2022MUL0001/_p
Copier
@ARTICLE{e106-d_1_31,
author={Renya DAIMO, Satoshi ONO, },
journal={IEICE TRANSACTIONS on Information},
title={Projection-Based Physical Adversarial Attack for Monocular Depth Estimation},
year={2023},
volume={E106-D},
number={1},
pages={31-35},
abstract={Monocular depth estimation has improved drastically due to the development of deep neural networks (DNNs). However, recent studies have revealed that DNNs for monocular depth estimation contain vulnerabilities that can lead to misestimation when perturbations are added to input. This study investigates whether DNNs for monocular depth estimation is vulnerable to misestimation when patterned light is projected on an object using a video projector. To this end, this study proposes an evolutionary adversarial attack method with multi-fidelity evaluation scheme that allows creating adversarial examples under black-box condition while suppressing the computational cost. Experiments in both simulated and real scenes showed that the designed light pattern caused a DNN to misestimate objects as if they have moved to the back.},
keywords={},
doi={10.1587/transinf.2022MUL0001},
ISSN={1745-1361},
month={January},}
Copier
TY - JOUR
TI - Projection-Based Physical Adversarial Attack for Monocular Depth Estimation
T2 - IEICE TRANSACTIONS on Information
SP - 31
EP - 35
AU - Renya DAIMO
AU - Satoshi ONO
PY - 2023
DO - 10.1587/transinf.2022MUL0001
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E106-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 2023
AB - Monocular depth estimation has improved drastically due to the development of deep neural networks (DNNs). However, recent studies have revealed that DNNs for monocular depth estimation contain vulnerabilities that can lead to misestimation when perturbations are added to input. This study investigates whether DNNs for monocular depth estimation is vulnerable to misestimation when patterned light is projected on an object using a video projector. To this end, this study proposes an evolutionary adversarial attack method with multi-fidelity evaluation scheme that allows creating adversarial examples under black-box condition while suppressing the computational cost. Experiments in both simulated and real scenes showed that the designed light pattern caused a DNN to misestimate objects as if they have moved to the back.
ER -