TY - GEN
T1 - All You Need Is RAW
T2 - 17th European Conference on Computer Vision, ECCV 2022
AU - Zhang, Yuxuan
AU - Dong, Bo
AU - Heide, Felix
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - Existing neural networks for computer vision tasks are vulnerable to adversarial attacks: adding imperceptible perturbations to the input images can fool these models into making a false prediction on an image that was correctly predicted without the perturbation. Various defense methods have proposed image-to-image mapping methods, either including these perturbations in the training process or removing them in a preprocessing step. In doing so, existing methods often ignore that the natural RGB images in today’s datasets are not captured but, in fact, recovered from RAW color filter array captures that are subject to various degradations in the capture. In this work, we exploit this RAW data distribution as an empirical prior for adversarial defense. Specifically, we propose a model-agnostic adversarial defensive method, which maps the input RGB images to Bayer RAW space and back to output RGB using a learned camera image signal processing (ISP) pipeline to eliminate potential adversarial patterns. The proposed method acts as an off-the-shelf preprocessing module and, unlike model-specific adversarial training methods, does not require adversarial images to train. As a result, the method generalizes to unseen tasks without additional retraining. Experiments on large-scale datasets, e.g., ImageNet, COCO, for different vision tasks, e.g., classification, semantic segmentation, object detection, validate that the method significantly outperforms existing methods across task domains.
AB - Existing neural networks for computer vision tasks are vulnerable to adversarial attacks: adding imperceptible perturbations to the input images can fool these models into making a false prediction on an image that was correctly predicted without the perturbation. Various defense methods have proposed image-to-image mapping methods, either including these perturbations in the training process or removing them in a preprocessing step. In doing so, existing methods often ignore that the natural RGB images in today’s datasets are not captured but, in fact, recovered from RAW color filter array captures that are subject to various degradations in the capture. In this work, we exploit this RAW data distribution as an empirical prior for adversarial defense. Specifically, we propose a model-agnostic adversarial defensive method, which maps the input RGB images to Bayer RAW space and back to output RGB using a learned camera image signal processing (ISP) pipeline to eliminate potential adversarial patterns. The proposed method acts as an off-the-shelf preprocessing module and, unlike model-specific adversarial training methods, does not require adversarial images to train. As a result, the method generalizes to unseen tasks without additional retraining. Experiments on large-scale datasets, e.g., ImageNet, COCO, for different vision tasks, e.g., classification, semantic segmentation, object detection, validate that the method significantly outperforms existing methods across task domains.
KW - Adversarial defense
KW - Low-level imaging
KW - Neural image processing
UR - http://www.scopus.com/inward/record.url?scp=85142681050&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85142681050&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-19800-7_19
DO - 10.1007/978-3-031-19800-7_19
M3 - Conference contribution
AN - SCOPUS:85142681050
SN - 9783031197994
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 323
EP - 343
BT - Computer Vision – ECCV 2022 - 17th European Conference, Proceedings
A2 - Avidan, Shai
A2 - Brostow, Gabriel
A2 - Cissé, Moustapha
A2 - Farinella, Giovanni Maria
A2 - Hassner, Tal
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 23 October 2022 through 27 October 2022
ER -