liu.seSearch for publications in DiVA
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Unsupervised Adversarial Learning of Anomaly Detection in the Wild
Linköping University, Department of Electrical Engineering, Computer Vision. Linköping University, Faculty of Science & Engineering. Termisk Systemteknik AB.ORCID iD: 0000-0002-6591-9400
Linköping University, Department of Electrical Engineering, Computer Vision. Linköping University, Faculty of Science & Engineering. Termisk Systemteknik AB.ORCID iD: 0000-0002-6763-5487
Linköping University, Department of Electrical Engineering, Computer Vision. Linköping University, Faculty of Science & Engineering.ORCID iD: 0000-0002-6096-3648
2020 (English)In: Proceedings of the 24th European Conference on Artificial Intelligence (ECAI) / [ed] Giuseppe De Giacomo, Alejandro Catala, Bistra Dilkina, Michela Milano, Senén Barro, Alberto Bugarín, Jérôme Lang, Amsterdam: IOS Press, 2020, Vol. 325, p. 1002-1008Conference paper, Published paper (Refereed)
Abstract [en]

Unsupervised learning of anomaly detection in high-dimensional data, such as images, is a challenging problem recently subject to intense research. Through careful modelling of the data distribution of normal samples, it is possible to detect deviant samples, so called anomalies. Generative Adversarial Networks (GANs) can model the highly complex, high-dimensional data distribution of normal image samples, and have shown to be a suitable approach to the problem. Previously published GAN-based anomaly detection methods often assume that anomaly-free data is available for training. However, this assumption is not valid in most real-life scenarios, a.k.a. in the wild. In this work, we evaluate the effects of anomaly contaminations in the training data on state-of-the-art GAN-based anomaly detection methods. As expected, detection performance deteriorates. To address this performance drop, we propose to add an additional encoder network already at training time and show that joint generator-encoder training stratifies the latent space, mitigating the problem with contaminated data. We show experimentally that the norm of a query image in this stratified latent space becomes a highly significant cue to discriminate anomalies from normal data. The proposed method achieves state-of-the-art performance on CIFAR-10 as well as on a large, previously untested dataset with cell images.

Place, publisher, year, edition, pages
Amsterdam: IOS Press, 2020. Vol. 325, p. 1002-1008
Series
Frontiers in Artificial Intelligence and Applications, ISSN 0922-6389, E-ISSN 1879-8314 ; 325
Keywords [en]
anomaly detection, GANs, generative adversarial networks, deep learning
National Category
Computer Vision and Robotics (Autonomous Systems)
Identifiers
URN: urn:nbn:se:liu:diva-174310DOI: 10.3233/FAIA200194ISI: 000650971301032ISBN: 9781643681009 (print)ISBN: 9781643681016 (electronic)OAI: oai:DiVA.org:liu-174310DiVA, id: diva2:1539624
Conference
24th European Conference on Artificial Intelligence (ECAI)
Projects
Learning Systems for Remote ThermographyEnergy Minimization for Computational CamerasELLIITAggregate FARming in the CLOUD (AFarCloud)
Funder
Swedish Research Council, D0570301EU, Horizon 2020, 783221Swedish Research Council, 2014-6227ELLIIT - The Linköping‐Lund Initiative on IT and Mobile Communications
Note

Funding: Swedish Research Council through the project Learning Systems for Remote Thermography [D0570301]; Swedish Research Council through project Energy Minimization for Computational Cameras [2014-6227]; Swedish Research Council through project ELLIIT (the Strategic Area for ICT research - Swedish Government); European Unions Horizon 2020 reseach and innovation programme [783221]

Available from: 2021-03-24 Created: 2021-03-24 Last updated: 2022-06-16Bibliographically approved

Open Access in DiVA

fulltext(569 kB)472 downloads
File information
File name FULLTEXT01.pdfFile size 569 kBChecksum SHA-512
74363ee8c389b85727d07f73edb030f8018cdc6c7bb3546dc0f86195e30f437ad0fc980e3bb3f3040d73ce218f51470a34c21df3f8c57fc93638e8099c798f28
Type fulltextMimetype application/pdf

Other links

Publisher's full text

Authority records

Berg, AmandaAhlberg, JörgenFelsberg, Michael

Search in DiVA

By author/editor
Berg, AmandaAhlberg, JörgenFelsberg, Michael
By organisation
Computer VisionFaculty of Science & Engineering
Computer Vision and Robotics (Autonomous Systems)

Search outside of DiVA

GoogleGoogle Scholar
Total: 472 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 491 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf