MineGAN: Effective Knowledge Transfer From GANs to Target Domains With Few Images

↓ Direkt till sidans innehåll
↓ Direkt till sidans sekundära innehåll (sidomenyn)

Search: id:"swepub:oai:DiVA.org:liu-168124" > MineGAN :

1 of 1
Previous record
Next record
To hitlist

Details
MARC

Wang, Y.Universitat Autonoma de Barcelona, Spain (author)

MineGAN : Effective Knowledge Transfer From GANs to Target Domains With Few Images

Article/chapterEnglish2020

Publisher, publication year, extent ...

IEEE,2020
printrdacarrier

Numbers

LIBRIS-ID:oai:DiVA.org:liu-168124
https://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-168124URI
https://doi.org/10.1109/CVPR42600.2020.00935DOI

Supplementary language notes

Language:English
Summary in:English

Part of subdatabase

SwePubSwePub

Classification

Subject category:ref swepub-contenttype
Subject category:kon swepub-publicationtype

Notes

One of the attractive characteristics of deep neural networks is their ability to transfer knowledge obtained in one domain to other related domains. As a result, high-quality networks can be trained in domains with relatively little training data. This property has been extensively studied for discriminative networks but has received significantly less attention for generative models. Given the often enormous effort required to train GANs, both computationally as well as in the dataset collection, the re-use of pretrained GANs is a desirable objective. We propose a novel knowledge transfer method for generative models based on mining the knowledge that is most beneficial to a specific target domain, either from a single or multiple pretrained GANs. This is done using a miner network that identifies which part of the generative distribution of each pretrained GAN outputs samples closest to the target domain. Mining effectively steers GAN sampling towards suitable regions of the latent space, which facilitates the posterior finetuning and avoids pathologies of other methods such as mode collapse and lack of flexibility. We perform experiments on several complex datasets using various GAN architectures (BigGAN, Progressive GAN) and show that the proposed method, called MineGAN, effectively transfers knowledge to domains with few target images, outperforming existing methods. In addition, MineGAN can successfully transfer knowledge from multiple pretrained GANs. Our code is available at: https://github.com/yaxingwang/MineGAN.

Subject headings and genre

NATURVETENSKAP Data- och informationsvetenskap Datorseende och robotik hsv//swe
NATURAL SCIENCES Computer and Information Sciences Computer Vision and Robotics hsv//eng
Gallium nitride;Generators;Generative adversarial networks;Training;Data mining;Knowledge transfer;Computational modeling

Added entries (persons, corporate bodies, meetings, titles ...)

Gonzalez-Garcia, A.Universitat Autonoma de Barcelona, Spain (author)
Berga, D.Universitat Autonoma de Barcelona, Spain (author)
Herranz, L.Universitat Autonoma de Barcelona, Spain (author)
Khan, Fahad Shahbaz,1983-Linköpings universitet,Datorseende,Tekniska fakulteten,Inception Institute of Artificial Intelligence, UAE(Swepub:liu)fahkh30 (author)
Weijer, J. van deUniversitat Autonoma de Barcelona, Spain (author)
Universitat Autonoma de Barcelona, SpainDatorseende (creator_code:org_t)

Related titles

In:2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR): IEEE, s. 9329-93389781728171685

Internet link

Find in a library

2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (Search for host publication in LIBRIS)

To the university's database

1 of 1
Previous record
Next record
To hitlist

Find more in SwePub

By the author/editor: Wang, Y.; Gonzalez-Garcia, ...; Berga, D.; Herranz, L.; Khan, Fahad Shah ...; Weijer, J. van d ...

About the subject

NATURAL SCIENCES: NATURAL SCIENCES; and Computer and Inf ...; and Computer Vision ...

Articles in the publication: 2020 IEEE/CVF Co ...

By the university: Linköping University

Search outside SwePub

Extend your search to:: Google; Google Book Search; Google Scholar

Kungliga biblioteket hanterar dina personuppgifter i enlighet med EU:s dataskyddsförordning (2018), GDPR. Läs mer om hur det funkar här.
Så här hanterar KB dina uppgifter vid användning av denna tjänst.

LIBRIS.kb.se