A Study on Model Compression Methods for SRGAN

Dong Hwi Kim, Jun Won Lee, Sang Hyo Park

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

3 Scopus citations

Abstract

The construction of SR algorithms by using deep learning model such as super-resolution generative adversarial networks (SRGAN) have become larger and complicated model architectures with requiring a vast amount of memory capacity. However, it is difficult to operate deep learning models which have millions of parameters at the mobile devices. Thus, in this paper, we present a study on lightweight neural network using network pruning method. Through our extensive experiments, pruned network can show similar performance to the original SRGAN model with substantially reduced model size.

Original languageEnglish
Title of host publication2022 International Conference on Electronics, Information, and Communication, ICEIC 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665409346
DOIs
StatePublished - 2022
Event2022 International Conference on Electronics, Information, and Communication, ICEIC 2022 - Jeju, Korea, Republic of
Duration: 6 Feb 20229 Feb 2022

Publication series

Name2022 International Conference on Electronics, Information, and Communication, ICEIC 2022

Conference

Conference2022 International Conference on Electronics, Information, and Communication, ICEIC 2022
Country/TerritoryKorea, Republic of
CityJeju
Period6/02/229/02/22

Keywords

  • Deep learning
  • Fine-tuning
  • Knowledge-distillation
  • Lightweight neural network
  • Network compression
  • Network pruning
  • Super-resolution

Fingerprint

Dive into the research topics of 'A Study on Model Compression Methods for SRGAN'. Together they form a unique fingerprint.

Cite this