Maintaining Natural Image Statistics with the Contextual Loss

Roey Mechrez, Itamar Talmi, Firas Shama, Lihi Zelnik-Manor

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Maintaining natural image statistics is a crucial factor in restoration and generation of realistic looking images. When training CNNs, photorealism is usually attempted by adversarial training (GAN), that pushes the output images to lie on the manifold of natural images. GANs are very powerful, but not perfect. They are hard to train and the results still often suffer from artifacts. In this paper we propose a complementary approach, that could be applied with or without GAN, whose goal is to train a feed-forward CNN to maintain natural internal statistics. We look explicitly at the distribution of features in an image and train the network to generate images with natural feature distributions. Our approach reduces by orders of magnitude the number of images required for training and achieves state-of-the-art results on both single-image super-resolution, and high-resolution surface normal estimation. Project page: https://www.github.com/roimehrez/contextualLoss.

Original languageEnglish
Title of host publicationComputer Vision - ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers
EditorsHongdong Li, Greg Mori, Konrad Schindler, C.V. Jawahar
Pages427-443
Number of pages17
DOIs
StatePublished - 2019
Event14th Asian Conference on Computer Vision, ACCV 2018 - Perth, Australia
Duration: 2 Dec 20186 Dec 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11363 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th Asian Conference on Computer Vision, ACCV 2018
Country/TerritoryAustralia
CityPerth
Period2/12/186/12/18

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Maintaining Natural Image Statistics with the Contextual Loss'. Together they form a unique fingerprint.

Cite this