Name:
123540426.pdf
Size:
1.981Mb
Format:
PDF
Description:
Accepted Manuscript
Embargo End Date:
2021-11-05
Type
Conference PaperAuthors
Pérez, Juan C.Alfarra, Motasem
Jeanneret, Guillaume
Bibi, Adel

Thabet, Ali Kassem

Ghanem, Bernard

Arbeláez, Pablo
KAUST Department
Electrical EngineeringComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Electrical Engineering Program
Visual Computing Center (VCC)
KAUST Grant Number
OSR-CRG2019-4033Date
2020-11-05Embargo End Date
2021-11-05Permanent link to this record
http://hdl.handle.net/10754/666421
Metadata
Show full item recordAbstract
We revisit the benefits of merging classical vision concepts with deep learning models. In particular, we explore the effect of replacing the first layers of various deep architectures with Gabor layers (i.e. convolutional layers with filters that are based on learnable Gabor parameters) on robustness against adversarial attacks. We observe that architectures with Gabor layers gain a consistent boost in robustness over regular models and maintain high generalizing test performance. We then exploit the analytical expression of Gabor filters to derive a compact expression for a Lipschitz constant of such filters, and harness this theoretical result to develop a regularizer we use during training to further enhance network robustness. We conduct extensive experiments with various architectures (LeNet, AlexNet, VGG16, and WideResNet) on several datasets (MNIST, SVHN, CIFAR10 and CIFAR100) and demonstrate large empirical robustness gains. Furthermore, we experimentally show how our regularizer provides consistent robustness improvements.Citation
Pérez, J. C., Alfarra, M., Jeanneret, G., Bibi, A., Thabet, A., Ghanem, B., & Arbeláez, P. (2020). Gabor Layers Enhance Network Robustness. Lecture Notes in Computer Science, 450–466. doi:10.1007/978-3-030-58545-7_26Sponsors
This work was partially supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No. OSR-CRG2019-4033.Publisher
Springer International PublishingConference/Event name
16th European Conference on Computer Vision, ECCV 2020ISBN
9783030585440arXiv
1912.05661Additional Links
http://link.springer.com/10.1007/978-3-030-58545-7_26ae974a485f413a2113503eed53cd6c53
10.1007/978-3-030-58545-7_26