Conference item icon

Conference item

On the robustness of quality measures for GANs

Abstract:

This work evaluates the robustness of quality measures of generative models such as Inception Score (IS) and Fréchet Inception Distance (FID). Analogous to the vulnerability of deep models against a variety of adversarial attacks, we show that such metrics can also be manipulated by additive pixel perturbations. Our experiments indicate that one can generate a distribution of images with very high scores but low perceptual quality. Conversely, one can optimize for small imperceptible perturba...

Expand abstract
Publication status:
Published
Peer review status:
Peer reviewed

Actions


Access Document


Files:
Publisher copy:
10.1007/978-3-031-19790-1_2

Authors


More by this author
Institution:
University of Oxford
Division:
MPLS
Department:
Engineering Science
Role:
Author
Publisher:
Springer
Host title:
Proceedings of the 17th European Conference on Computer Vision (ECCV 2022)
Series:
Lecture Notes in Computer Science
Volume:
13677
Pages:
18-33
Publication date:
2022-10-24
Acceptance date:
2022-07-03
Event title:
17th European Conference on Computer Vision (ECCV 2022)
Event location:
Tel Aviv, Israel
Event website:
https://eccv2022.ecva.net/
Event start date:
2022-10-23
Event end date:
2022-10-27
DOI:
EISSN:
1611-3349
ISSN:
0302-9743
ISBN:
9783031197895
Language:
English
Keywords:
Pubs id:
1311534
Local pid:
pubs:1311534
Deposit date:
2022-12-12

Terms of use


Views and Downloads






If you are the owner of this record, you can report an update to it here: Report update to this record

TO TOP