Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX

Abstract: 

This paper presents a simple yet effective method to improve the visual quality of Generative Adversarial Network (GAN) generated images. In typical GAN architectures, the discriminator block is designed mainly to capture the class-specific content from images without explicitly imposing constraints on the visual quality of the generated images. A key insight from the image quality assessment literature is that natural scenes possess a very unique local structural and (hence) statistical signature, and that distortions affect this signature. We translate this insight into a constraint on the loss function of the discriminator in the GAN architecture with the goal of improving the visual quality of the generated images. Specifically, this constraint is based on the Multi-scale Structural Similarity (MS-SSIM) index to guarantee local structural and statistical integrity. We train GANs (Boundary Equilibrium GANs, to be precise,) using the proposed approach on popular face and car image databases and demonstrate the improvement relative to standard training approaches both visually and quantitatively.

up
0 users have voted:

Paper Details

Authors:
Submitted On:
7 October 2018 - 6:01am
Short Link:
Type:
Poster
Event:
Presenter's Name:
Sumohana S. Channappayya
Paper Code:
2632
Document Year:
2018
Cite

Document Files

ICIP_2018_BEGAN.pdf

(120)

Subscribe

[1] , "IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX", IEEE SigPort, 2018. [Online]. Available: http://sigport.org/3590. Accessed: Jul. 19, 2019.
@article{3590-18,
url = {http://sigport.org/3590},
author = { },
publisher = {IEEE SigPort},
title = {IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX},
year = {2018} }
TY - EJOUR
T1 - IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX
AU -
PY - 2018
PB - IEEE SigPort
UR - http://sigport.org/3590
ER -
. (2018). IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX. IEEE SigPort. http://sigport.org/3590
, 2018. IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX. Available at: http://sigport.org/3590.
. (2018). "IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX." Web.
1. . IMPROVING THE VISUAL QUALITY OF GENERATIVE ADVERSARIAL NETWORK (GAN)-GENERATED IMAGES USING THE MULTI-SCALE STRUCTURAL SIMILARITY INDEX [Internet]. IEEE SigPort; 2018. Available from : http://sigport.org/3590