Towards real-time interest point detection and description for mobile and robotic devices

Convolutional Neural Networks (CNNs) have been successfully adopted by state-of-the-art feature point detection and description networks for the past number of years. The focus of these systems has been predominately on the accuracy of the system, rather than on its efficiency or ability to be imple...

Full description

Saved in:
Bibliographic Details
Main Authors: Patrick Rowsome, Muhammad Adil Raja, R. Muhammad Atif Azad
Format: Article
Language:English
Published: KeAi Communications Co., Ltd. 2024-09-01
Series:Informatics and Health
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2949953424000134
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849690356387414016
author Patrick Rowsome
Muhammad Adil Raja
R. Muhammad Atif Azad
author_facet Patrick Rowsome
Muhammad Adil Raja
R. Muhammad Atif Azad
author_sort Patrick Rowsome
collection DOAJ
description Convolutional Neural Networks (CNNs) have been successfully adopted by state-of-the-art feature point detection and description networks for the past number of years. The focus of these systems has been predominately on the accuracy of the system, rather than on its efficiency or ability to be implemented in real-time on embedded robotic devices. This paper demonstrates how techniques, developed for other CNN use cases, can be integrated into interest point detection and description systems to compress their network size and reduce the computational complexity; this reduces the barrier to their uptake in computationally challenged environments. This paper documents the integration of these techniques into the popular Reliable Detector and Descriptor (R2D2) network. Along with the integration details, a comprehensive Key Performance Indicator (KPI) framework is developed to test all aspects of the networks. As a result, this paper presents a lightweight variant of the R2D2 network that significantly reduces parameters and computational complexity while crucially maintaining an acceptable level of accuracy. Consequently, this new compressed network is more appropriate for use in real world systems and advances the efforts to implement such CNN based system for mobile devices.
format Article
id doaj-art-8ae7ae59d03641d8a2be62c5f0901580
institution DOAJ
issn 2949-9534
language English
publishDate 2024-09-01
publisher KeAi Communications Co., Ltd.
record_format Article
series Informatics and Health
spelling doaj-art-8ae7ae59d03641d8a2be62c5f09015802025-08-20T03:21:19ZengKeAi Communications Co., Ltd.Informatics and Health2949-95342024-09-0112829210.1016/j.infoh.2024.06.002Towards real-time interest point detection and description for mobile and robotic devicesPatrick Rowsome0Muhammad Adil Raja1R. Muhammad Atif Azad2Protex AI, IrelandRegulated Software Research Center (RSRC), Dundalk Institute of Technology, Country Louth, Ireland; Corresponding author.School of Computing and Digital Technology, Birmingham City University, Birmingham, UKConvolutional Neural Networks (CNNs) have been successfully adopted by state-of-the-art feature point detection and description networks for the past number of years. The focus of these systems has been predominately on the accuracy of the system, rather than on its efficiency or ability to be implemented in real-time on embedded robotic devices. This paper demonstrates how techniques, developed for other CNN use cases, can be integrated into interest point detection and description systems to compress their network size and reduce the computational complexity; this reduces the barrier to their uptake in computationally challenged environments. This paper documents the integration of these techniques into the popular Reliable Detector and Descriptor (R2D2) network. Along with the integration details, a comprehensive Key Performance Indicator (KPI) framework is developed to test all aspects of the networks. As a result, this paper presents a lightweight variant of the R2D2 network that significantly reduces parameters and computational complexity while crucially maintaining an acceptable level of accuracy. Consequently, this new compressed network is more appropriate for use in real world systems and advances the efforts to implement such CNN based system for mobile devices.http://www.sciencedirect.com/science/article/pii/S2949953424000134Interest point detectionComputer visionConvolutional neural networksDeep learningReliable detector and descriptor
spellingShingle Patrick Rowsome
Muhammad Adil Raja
R. Muhammad Atif Azad
Towards real-time interest point detection and description for mobile and robotic devices
Informatics and Health
Interest point detection
Computer vision
Convolutional neural networks
Deep learning
Reliable detector and descriptor
title Towards real-time interest point detection and description for mobile and robotic devices
title_full Towards real-time interest point detection and description for mobile and robotic devices
title_fullStr Towards real-time interest point detection and description for mobile and robotic devices
title_full_unstemmed Towards real-time interest point detection and description for mobile and robotic devices
title_short Towards real-time interest point detection and description for mobile and robotic devices
title_sort towards real time interest point detection and description for mobile and robotic devices
topic Interest point detection
Computer vision
Convolutional neural networks
Deep learning
Reliable detector and descriptor
url http://www.sciencedirect.com/science/article/pii/S2949953424000134
work_keys_str_mv AT patrickrowsome towardsrealtimeinterestpointdetectionanddescriptionformobileandroboticdevices
AT muhammadadilraja towardsrealtimeinterestpointdetectionanddescriptionformobileandroboticdevices
AT rmuhammadatifazad towardsrealtimeinterestpointdetectionanddescriptionformobileandroboticdevices