Adversarial training for sketch retrieval
File(s)
Accepted version
Author(s)
Creswell, A
Bharath, AA
Type
Conference Paper
Abstract
Generative Adversarial Networks (GAN) are able to learn excellent
representations for unlabelled data which can be applied to image generation
and scene classification. Representations learned by GANs have not yet been
applied to retrieval. In this paper, we show that the representations learned
by GANs can indeed be used for retrieval. We consider heritage documents that
contain unlabelled Merchant Marks, sketch-like symbols that are similar to
hieroglyphs. We introduce a novel GAN architecture with design features that
make it suitable for sketch retrieval. The performance of this sketch-GAN is
compared to a modified version of the original GAN architecture with respect to
simple invariance properties. Experiments suggest that sketch-GANs learn
representations that are suitable for retrieval and which also have increased
stability to rotation, scale and translation compared to the standard GAN
architecture.
representations for unlabelled data which can be applied to image generation
and scene classification. Representations learned by GANs have not yet been
applied to retrieval. In this paper, we show that the representations learned
by GANs can indeed be used for retrieval. We consider heritage documents that
contain unlabelled Merchant Marks, sketch-like symbols that are similar to
hieroglyphs. We introduce a novel GAN architecture with design features that
make it suitable for sketch retrieval. The performance of this sketch-GAN is
compared to a modified version of the original GAN architecture with respect to
simple invariance properties. Experiments suggest that sketch-GANs learn
representations that are suitable for retrieval and which also have increased
stability to rotation, scale and translation compared to the standard GAN
architecture.
Date Issued
2016-09-18
Date Acceptance
2016-09-01
Citation
Lecture Notes in Computer Science, 2016, 9913
ISBN
978-3-319-46603-3
ISSN
0302-9743
Publisher
Springer Verlag
Journal / Book Title
Lecture Notes in Computer Science
Volume
9913
Copyright Statement
© Springer International Publishing Switzerland 2016. The final publication is available at Springer via https://link.springer.com/chapter/10.1007%2F978-3-319-46604-0_55
Identifier
http://arxiv.org/abs/1607.02748v2
Source
Computer Vision – ECCV 2016 Workshops
Subjects
cs.CV
cs.CV
Notes
Accepted to ECCV2016 VisArt Workshop
Start Date
2016-10-08
Finish Date
2016-10-16
Coverage Spatial
Amsterdam, The Netherlands