Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • 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
Improving Image Autoencoder Embeddings with Perceptual Loss
Luleå University of Technology, Department of Computer Science, Electrical and Space Engineering, Embedded Internet Systems Lab.ORCID iD: 0000-0003-0100-4030
Luleå University of Technology, Department of Computer Science, Electrical and Space Engineering, Embedded Internet Systems Lab.ORCID iD: 0000-0001-5662-825X
Luleå University of Technology, Department of Computer Science, Electrical and Space Engineering, Embedded Internet Systems Lab.ORCID iD: 0000-0003-4029-6574
2020 (English)In: 2020 International Joint Conference on Neural Networks (IJCNN), IEEE, 2020, article id 20229Conference paper, Published paper (Refereed)
Abstract [en]

Autoencoders are commonly trained using element-wise loss. However, element-wise loss disregards high-level structures in the image which can lead to embeddings that disregard them as well. A recent improvement to autoencoders that helps alleviate this problem is the use of perceptual loss. This work investigates perceptual loss from the perspective of encoder embeddings themselves. Autoencoders are trained to embed images from three different computer vision datasets using perceptual loss based on a pretrained model as well as pixel-wise loss. A host of different predictors are trained to perform object positioning and classification on the datasets given the embedded images as input. The two kinds of losses are evaluated by comparing how the predictors performed with embeddings from the differently trained autoencoders. The results show that, in the image domain, the embeddings generated by autoencoders trained with perceptual loss enable more accurate predictions than those trained with element-wise loss. Furthermore, the results show that, on the task of object positioning of a small-scale feature, perceptual loss can improve the results by a factor 10. The experimental setup is available online: https://github.com/guspih/Perceptual-Autoencoders

Place, publisher, year, edition, pages
IEEE, 2020. article id 20229
Series
International Joint Conference on Neural Networks (IJCNN), ISSN 2161-4393, E-ISSN 2161-4407
Keywords [en]
Autoencoder, Perceptual Loss, Image Processing, Image Classification, Object Positioning, Embeddings
National Category
Other Electrical Engineering, Electronic Engineering, Information Engineering Computer Sciences
Research subject
Machine Learning; Cyber-Physical Systems
Identifiers
URN: urn:nbn:se:ltu:diva-80609DOI: 10.1109/IJCNN48605.2020.9207431ISI: 000626021406074Scopus ID: 2-s2.0-85093844245OAI: oai:DiVA.org:ltu-80609DiVA, id: diva2:1462523
Conference
2020 International Joint Conference on Neural Networks (IJCNN), 18-24 July, 2020, Glasgow, United Kingdom
Note

ISBN för värdpublikation: 978-1-7281-6926-2, 978-1-7281-6927-9

Available from: 2020-08-31 Created: 2020-08-31 Last updated: 2025-10-22Bibliographically approved
In thesis
1. Deep Perceptual Loss for Improved Downstream Prediction
Open this publication in new window or tab >>Deep Perceptual Loss for Improved Downstream Prediction
2021 (English)Licentiate thesis, comprehensive summary (Other academic)
Place, publisher, year, edition, pages
Luleå: Luleå tekniska universitet, 2021
Series
Licentiate thesis / Luleå University of Technology, ISSN 1402-1757
National Category
Computer graphics and computer vision Computer Sciences
Research subject
Machine Learning
Identifiers
urn:nbn:se:ltu:diva-86440 (URN)978-91-7790-904-0 (ISBN)978-91-7790-905-7 (ISBN)
Presentation
2021-10-12, C305, Luleå Tekniska Universitet, Luleå, 14:00 (English)
Opponent
Supervisors
Available from: 2021-08-16 Created: 2021-08-13 Last updated: 2025-10-21Bibliographically approved

Open Access in DiVA

fulltext(362 kB)767 downloads
File information
File name FULLTEXT01.pdfFile size 362 kBChecksum SHA-512
994c4a396f4412aa37ff2d3291296ffca4d403cbf5b42252053876c56b6872fe94d5aa1ec6c6a515736369e316018f3d71cde7cf8ee7575577df4225a6f90f33
Type fulltextMimetype application/pdf

Other links

Publisher's full textScopus

Authority records

Grund Pihlgren, GustavSandin, FredrikLiwicki, Marcus

Search in DiVA

By author/editor
Grund Pihlgren, GustavSandin, FredrikLiwicki, Marcus
By organisation
Embedded Internet Systems Lab
Other Electrical Engineering, Electronic Engineering, Information EngineeringComputer Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 769 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
urn-nbn

Altmetric score

doi
urn-nbn
Total: 653 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • 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