Show simple item record

dc.contributor.authorAlfarra, Motasem
dc.contributor.authorBibi, Adel
dc.contributor.authorKhan, Naeemullah
dc.contributor.authorTorr, Philip H. S.
dc.contributor.authorGhanem, Bernard
dc.date.accessioned2021-07-14T06:39:22Z
dc.date.available2021-07-14T06:39:22Z
dc.date.issued2021-07-02
dc.identifier.urihttp://hdl.handle.net/10754/670197
dc.description.abstractDeep neural networks are vulnerable to input deformations in the form of vector fields of pixel displacements and to other parameterized geometric deformations e.g. translations, rotations, etc. Current input deformation certification methods either (i) do not scale to deep networks on large input datasets, or (ii) can only certify a specific class of deformations, e.g. only rotations. We reformulate certification in randomized smoothing setting for both general vector field and parameterized deformations and propose DeformRS-VF and DeformRS-Par, respectively. Our new formulation scales to large networks on large input datasets. For instance, DeformRS-Par certifies rich deformations, covering translations, rotations, scaling, affine deformations, and other visually aligned deformations such as ones parameterized by Discrete-Cosine-Transform basis. Extensive experiments on MNIST, CIFAR10 and ImageNet show that DeformRS-Par outperforms existing state-of-the-art in certified accuracy, e.g. improved certified accuracy of 6% against perturbed rotations in the set [-10,10] degrees on ImageNet.
dc.description.sponsorshipThis work was partially supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research.
dc.publisherarXiv
dc.relation.urlhttps://arxiv.org/pdf/2107.00996.pdf
dc.rightsArchived with thanks to arXiv
dc.titleDeformRS: Certifying Input Deformations with Randomized Smoothing
dc.typePreprint
dc.contributor.departmentComputer, Electrical and Mathematical Science and Engineering (CEMSE) Division
dc.contributor.departmentElectrical and Computer Engineering
dc.contributor.departmentElectrical and Computer Engineering Program
dc.contributor.departmentVCC Analytics Research Group
dc.eprint.versionPre-print
dc.contributor.institutionUniversity of Oxford
dc.identifier.arxivid2107.00996
kaust.personAlfarra, Motasem
kaust.personGhanem, Bernard
dc.relation.issupplementedbygithub:MotasemAlfarra/DeformRS
refterms.dateFOA2021-07-14T06:39:55Z
display.relations<b>Is Supplemented By:</b><br/> <ul><li><i>[Software]</i> <br/> Title: MotasemAlfarra/DeformRS:. Publication Date: 2021-06-07. github: <a href="https://github.com/MotasemAlfarra/DeformRS" >MotasemAlfarra/DeformRS</a> Handle: <a href="http://hdl.handle.net/10754/670269" >10754/670269</a></a></li></ul>
kaust.acknowledged.supportUnitOffice of Sponsored Research


Files in this item

Thumbnail
Name:
Preprintfile1.pdf
Size:
3.431Mb
Format:
PDF
Description:
Pre-print

This item appears in the following Collection(s)

Show simple item record