Show simple item record

dc.contributor.authorHamdi, Abdullah
dc.contributor.authorGhanem, Bernard
dc.date.accessioned2021-01-21T08:37:43Z
dc.date.available2019-12-18T10:31:34Z
dc.date.available2021-01-21T08:37:43Z
dc.date.issued2021-01-10
dc.identifier.citationHamdi, A., & Ghanem, B. (2020). Towards Analyzing Semantic Robustness of Deep Neural Networks. Lecture Notes in Computer Science, 22–38. doi:10.1007/978-3-030-66415-2_2
dc.identifier.isbn9783030664145
dc.identifier.isbn9783030664152
dc.identifier.issn0302-9743
dc.identifier.issn1611-3349
dc.identifier.doi10.1007/978-3-030-66415-2_2
dc.identifier.urihttp://hdl.handle.net/10754/660663
dc.description.abstractDespite the impressive performance of Deep Neural Networks (DNNs) on various vision tasks, they still exhibit erroneous high sensitivity toward semantic primitives (e.g. object pose). We propose a theoretically grounded analysis for DNN robustness in the semantic space. We qualitatively analyze different DNNs’ semantic robustness by visualizing the DNN global behavior as semantic maps and observe interesting behavior of some DNNs. Since generating these semantic maps does not scale well with the dimensionality of the semantic space, we develop a bottom-up approach to detect robust regions of DNNs. To achieve this, we formalize the problem of finding robust semantic regions of the network as optimizing integral bounds and we develop expressions for update directions of the region bounds. We use our developed formulations to quantitatively evaluate the semantic robustness of different popular network architectures. We show through extensive experimentation that several networks, while trained on the same dataset and enjoying comparable accuracy, do not necessarily perform similarly in semantic robustness. For example, InceptionV3 is more accurate despite being less semantically robust than ResNet50. We hope that this tool will serve as a milestone towards understanding the semantic robustness of DNNs.
dc.description.sponsorshipThis work was supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research under Award No. OSR-CRG2018-3730
dc.publisherSpringer Nature
dc.relation.urlhttp://link.springer.com/10.1007/978-3-030-66415-2_2
dc.rightsArchived with thanks to Springer International Publishing
dc.titleTowards Analyzing Semantic Robustness of Deep Neural Networks
dc.typeConference Paper
dc.contributor.departmentElectrical Engineering Program
dc.contributor.departmentPhysical Science and Engineering (PSE) Division
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
dc.conference.dateAugust 2020
dc.conference.nameEuropean Conference on Computer Vision (ECCV) 2020
dc.conference.locationOnline
dc.eprint.versionPost-print
dc.identifier.pages22-38
dc.identifier.arxivid1904.04621
kaust.personHamdi, Abdullah
kaust.personGhanem, Bernard
kaust.grant.numberOSR-CRG2018-3730
dc.date.accepted2020-08-23
dc.relation.issupplementedbygithub:ajhamdi/semantic-robustness
refterms.dateFOA2019-12-18T10:32:35Z
display.relations<b>Is Supplemented By:</b><br/> <ul><li><i>[Software]</i> <br/> Title: ajhamdi/semantic-robustness: implementation of the paper: "Towards Analyzing Semantic Robustness of Deep Neural Networks" (ECCV 2020 workshop). Publication Date: 2019-03-20. github: <a href="https://github.com/ajhamdi/semantic-robustness" >ajhamdi/semantic-robustness</a> Handle: <a href="http://hdl.handle.net/10754/667397" >10754/667397</a></a></li></ul>
kaust.acknowledged.supportUnitOffice of Sponsored Research


Files in this item

Thumbnail
Name:
1904.04621.pdf
Size:
8.312Mb
Format:
PDF
Description:
Accepted Manuscript

This item appears in the following Collection(s)

Show simple item record

VersionItemEditorDateSummary

*Selected version