Show simple item record

dc.contributor.authorPérez, Juan C.
dc.contributor.authorAlfarra, Motasem
dc.contributor.authorJeanneret, Guillaume
dc.contributor.authorBibi, Adel
dc.contributor.authorThabet, Ali Kassem
dc.contributor.authorGhanem, Bernard
dc.contributor.authorArbeláez, Pablo
dc.date.accessioned2020-12-16T13:29:27Z
dc.date.available2020-12-16T13:29:27Z
dc.date.issued2020-11-05
dc.identifier.citationPérez, 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_26
dc.identifier.isbn9783030585440
dc.identifier.issn1611-3349
dc.identifier.issn0302-9743
dc.identifier.doi10.1007/978-3-030-58545-7_26
dc.identifier.urihttp://hdl.handle.net/10754/666421
dc.description.abstractWe 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.
dc.description.sponsorshipThis 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.
dc.publisherSpringer International Publishing
dc.relation.urlhttp://link.springer.com/10.1007/978-3-030-58545-7_26
dc.relation.urlhttp://arxiv.org/pdf/1912.05661
dc.rightsArchived with thanks to Springer International Publishing
dc.rightsThis file is an open access version redistributed from: http://arxiv.org/pdf/1912.05661
dc.titleGabor Layers Enhance Network Robustness
dc.typeConference Paper
dc.contributor.departmentElectrical Engineering
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
dc.contributor.departmentElectrical Engineering Program
dc.contributor.departmentVisual Computing Center (VCC)
dc.rights.embargodate2021-11-05
dc.conference.date2020-08-23 to 2020-08-28
dc.conference.name16th European Conference on Computer Vision, ECCV 2020
dc.conference.locationGlasgow, GBR
dc.eprint.versionPre-print
dc.contributor.institutionCenter for Research and Formation in Artificial Intelligence, Universidad de los Andes, Bogota, Colombia
dc.identifier.volume12354 LNCS
dc.identifier.pages450-466
dc.identifier.arxivid1912.05661
kaust.personAlfarra, Motasem
kaust.personBibi, Adel
kaust.personThabet, Ali Kassem
kaust.personGhanem, Bernard
kaust.grant.numberOSR-CRG2019-4033
dc.identifier.eid2-s2.0-85097093633
kaust.acknowledged.supportUnitOffice of Sponsored Research (OSR)


Files in this item

Thumbnail
Name:
123540426.pdf
Size:
1.981Mb
Format:
PDF
Description:
Accepted Manuscript
Embargo End Date:
2021-11-05

This item appears in the following Collection(s)

Show simple item record