elib
DLR-Header
DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Look ATME: The Discriminator Mean Entropy Needs Attention

Solano Carrillo, Edgardo and Bueno Rodriguez, Angel and Carrillo Perez, Borja Jesus and Steiniger, Yannik and Stoppe, Jannis (2023) Look ATME: The Discriminator Mean Entropy Needs Attention. In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023. CVPR 2023 Workshop on Generative Models in Computer Vision, Vancouver, Canada. doi: 10.1109/CVPRW59228.2023.00086. ISBN 979-835030129-8. ISSN 1063-6919.

[img] PDF
5MB

Abstract

Generative adversarial networks (GANs) are successfully used for image synthesis but are known to face instability during training. In contrast, probabilistic diffusion models (DMs) are stable and generate high-quality images, at the cost of an expensive sampling procedure. In this paper, we introduce a simple method to allow GANs to stably converge to their theoretical optimum, while bringing in the denoising machinery from DMs. These models are combined into a simpler model (ATME) that only requires a forward pass during inference, making predictions cheaper and more accurate than DMs and popular GANs. ATME breaks an information asymmetry existing in most GAN models in which the discriminator has spatial knowledge of where the generator is failing. To restore the information symmetry, the generator is endowed with knowledge of the entropic state of the discriminator, which is leveraged to allow the adversarial game to converge towards equilibrium. We demonstrate the power of our method in several image-to-image translation tasks, showing superior performance than state-of-the-art methods at a lesser cost. Code is available at https://github.com/DLR-MI/atme.

Item URL in elib:https://elib.dlr.de/195910/
Document Type:Conference or Workshop Item (Poster)
Title:Look ATME: The Discriminator Mean Entropy Needs Attention
Authors:
AuthorsInstitution or Email of AuthorsAuthor's ORCID iDORCID Put Code
Solano Carrillo, EdgardoUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Bueno Rodriguez, AngelUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Carrillo Perez, Borja JesusUNSPECIFIEDUNSPECIFIEDUNSPECIFIED
Steiniger, YannikUNSPECIFIEDhttps://orcid.org/0000-0002-9327-446XUNSPECIFIED
Stoppe, JannisUNSPECIFIEDhttps://orcid.org/0000-0003-2952-3422UNSPECIFIED
Date:2023
Journal or Publication Title:2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023
Refereed publication:Yes
Open Access:Yes
Gold Open Access:No
In SCOPUS:Yes
In ISI Web of Science:Yes
DOI:10.1109/CVPRW59228.2023.00086
ISSN:1063-6919
ISBN:979-835030129-8
Status:Published
Keywords:Deep learning, GAN, pix2pix, image-to-image translation, diffusion models
Event Title:CVPR 2023 Workshop on Generative Models in Computer Vision
Event Location:Vancouver, Canada
Event Type:international Conference
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Transport
HGF - Program Themes:other
DLR - Research area:Transport
DLR - Program:V - no assignment
DLR - Research theme (Project):V - no assignment
Location: Bremerhaven
Institutes and Institutions:Institute for the Protection of Maritime Infrastructures
Deposited By: Solano Carrillo, Edgardo
Deposited On:20 Nov 2023 14:00
Last Modified:20 Nov 2023 14:00

Repository Staff Only: item control page

Browse
Search
Help & Contact
Information
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.