Super-resolution with generative adversarial networks for improved object detection in aerial images
| Date | 21 November 2022 |
| Pages | 349-357 |
| DOI | https://doi.org/10.1108/IDD-05-2022-0048 |
| Published date | 21 November 2022 |
| Subject Matter | Library & information science,Library & information services,Lending,Document delivery,Collection building & management,Stock revision,Consortia |
| Author | Aslan Ahmet Haykir,Ilkay Oksuz |
Super-resolution with generative adversarial
networks for improved object detection in
aerial images
Aslan Ahmet Haykir and Ilkay Oksuz
Department of Computer Engineering, Istanbul Technical University, Istanbul, Turkey
Abstract
Purpose –Data quality and data resolution are essential for computer vision tasks like medical image processing, object detection, pattern recogni tion and so
on. Super-resolution is a way to increase the image resolution, and super-resolved images contain more information compared to their l ow-resolution
counterparts. The purpose of this study is analyzing the effects of the super resolution models trained before on object detection for aerialimages.
Design/methodology/approach –Two different models were trained using the Super-Resolution Generative Adversarial Network (SRGA N)
architecture on two aerial image data sets, the xView and the Dataset for Object deTection in Aerial images (DOTA). This study uses these models to
increase the resolution of aerial images for improving object detection performance. This study analyzes the effects of the model with the best
perceptual index (PI) and the model with the best RMSE on object detection in detail.
Findings –Super-resolution increases the object detection quality as expected. But, the super-resolution model wit h better perceptual quality
achieves lower mean average precision results compared to the model with better RMSE. It means that the model with a better PI is more
meaningful to human perception but less meaningful to computer vision.
Originality/value –The contributions of the authors to the literature are threefold. First, they do a wide analysis of SRGAN results for aerial image
super-resolution on the task of object detection. Second, they compare super-resolution models with best PI and best RMSE to showcase the
differences on object detection performance as a downstream task first time in the literature. Finally, they use a transfer learn ing approach for
super-resolution to improve the performance of object detection.
Keywords Data quality, Aerial images, Super-resolution, Object detection, Generative adversarial networks, Perceptual quality
Paper type Research paper
1. Introduction
Object detection from aerialimages is a challenging task due to
the quality of the images. The established object detection
algorithms can perform poorly, when original data is of poor
quality similar to general machine learning applications
(Sambasivan et al.,2021). The low-quality data needs to be
improved to be able to be used in downstream machine
learning tasks (e.g.object detection) (Chen et al.,2021).
Super-resolution is a method that can generate extra information
using a low-resolution image and increase the image resolution. A
super-resolution model can be trained using original images
captured by an imaging source with their low-resolution
counterparts. A super-resolved image is more meaningful to human
perception compared to its low-resolution version. Also, a super-
resolved image has advantages for computer vision tasks like object
detection and pattern recognition, because it increases the data
quality and enables more information about the original scenes.
Super-resolution can be used as a preprocessing method to improve
the performance of a computer vision task. To create a super-
resolution counterpart of the low-resolution image, generative
adversarial networks are commonly used. For example, Super-
Resolution Generative Adversarial Network (SRGAN) (Ledig
et al., 2016) is a generative adversarial network (GAN)-based
super-resolution architecture that we use to improve the
performance of object detection on aerial images. The difference of
the super-resolution compared to the bicubic interpolation along
with the original image can be seen in Figure 1.
Because low-quality data can cause object detection algorithms
perform poorly, the super-resolution can be used to improve data
quality. In this work, to solve the low data quality problem and
improve performance we use the super-resolution to improv e data
quality for object detection in aerial images. Previous works showed
that the super-resolution can be a good choice to improve object
detection performance. Different from some related works, we
used the SRGAN for the super-resolution because it is GAN based
The current issue and full text archiveof this journal is available on Emerald
Insight at: https://www.emerald.com/insight/2398-6247.htm
Information Discovery and Delivery
51/4 (2023) 349–357
© Emerald Publishing Limited [ISSN 2398-6247]
[DOI 10.1108/IDD-05-2022-0048]
This paper part of special section “Information and data quality for
intelligent systems”, guest edited by Junhua Ding, Haihua Chen, Lei Li
and Ismini Lourentzou.
This paper has been produced benefiting from the 2232 International
Fellowship for Outstanding Researchers Program of TUBITAK (Project
No. 118C353). However, the entire responsibility of the publication/paper
belongs to the owner of the paper. The financial support received from
TUBITAK does not mean that the content of the publication is approved
in a scientific sense by TUBITAK.
Received 30 May 2022
Revised 23 August 2022
18 October 2022
Accepted 19 October 2022
349
Get this document and AI-powered insights with a free trial of vLex and Vincent AI
Get Started for FreeStart Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
