Show simple item record

dc.contributor.authorSkorokhodov, Ivan
dc.contributor.authorIgnatyev, Savva
dc.contributor.authorElhoseiny, Mohamed
dc.date.accessioned2020-11-26T12:37:20Z
dc.date.available2020-11-26T12:37:20Z
dc.date.issued2020-11-24
dc.identifier.urihttp://hdl.handle.net/10754/666125
dc.description.abstractIn most existing learning systems, images are typically viewed as 2D pixel arrays. However, in another paradigm gaining popularity, a 2D image is represented as an implicit neural representation (INR) -- an MLP that predicts an RGB pixel value given its (x,y) coordinate. In this paper, we propose two novel architectural techniques for building INR-based image decoders: factorized multiplicative modulation and multi-scale INRs, and use them to build a state-of-the-art continuous image GAN. Previous attempts to adapt INRs for image generation were limited to MNIST-like datasets and do not scale to complex real-world data. Our proposed architectural design improves the performance of continuous image generators by x6-40 times and reaches FID scores of 6.27 on LSUN bedroom 256x256 and 16.32 on FFHQ 1024x1024, greatly reducing the gap between continuous image GANs and pixel-based ones. To the best of our knowledge, these are the highest reported scores for an image generator, that consists entirely of fully-connected layers. Apart from that, we explore several exciting properties of INR-based decoders, like out-of-the-box superresolution, meaningful image-space interpolation, accelerated inference of low-resolution images, an ability to extrapolate outside of image boundaries and strong geometric prior. The source code is available at https://github.com/universome/inr-gan
dc.publisherarXiv
dc.relation.urlhttps://arxiv.org/pdf/2011.12026
dc.rightsArchived with thanks to arXiv
dc.titleAdversarial Generation of Continuous Images
dc.typePreprint
dc.contributor.departmentKAUST Thuwal, Saudi Arabia.
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
dc.eprint.versionPre-print
dc.contributor.institutionSkoltech Moscow, Russia.
dc.identifier.arxivid2011.12026
kaust.personSkorokhodov, Ivan
kaust.personElhoseiny, Mohamed
dc.relation.issupplementedbygithub:universome/inr-gan
refterms.dateFOA2020-11-26T12:38:26Z
display.relations<b>Is Supplemented By:</b><br/> <ul><li><i>[Software]</i> <br/> Title: universome/inr-gan: Adversarial Generation of Continuous Images. Publication Date: 2020-11-24. github: <a href="https://github.com/universome/inr-gan" >universome/inr-gan</a> Handle: <a href="http://hdl.handle.net/10754/668008" >10754/668008</a></a></li></ul>


Files in this item

Thumbnail
Name:
Preprintfile1.pdf
Size:
6.657Mb
Format:
PDF
Description:
Pre-print

This item appears in the following Collection(s)

Show simple item record