MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection

<p dir="ltr">Accurate detection of pedestrian lanes is a crucial criterion for vision-impaired people to navigate freely and safely. The current deep learning methods have achieved reasonable accuracy at this task. However, they lack practicality for real-time pedestrian lane detecti...

Full description

Saved in:
Bibliographic Details
Main Author: Sui Paul Ang (18460605) (author)
Other Authors: Son Lam Phung (18460602) (author), Soan T. M. Duong (18460599) (author), Abdesselam Bouzerdoum (17900021) (author)
Published: 2023
Subjects:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1864513507679010816
author Sui Paul Ang (18460605)
author2 Son Lam Phung (18460602)
Soan T. M. Duong (18460599)
Abdesselam Bouzerdoum (17900021)
author2_role author
author
author
author_facet Sui Paul Ang (18460605)
Son Lam Phung (18460602)
Soan T. M. Duong (18460599)
Abdesselam Bouzerdoum (17900021)
author_role author
dc.creator.none.fl_str_mv Sui Paul Ang (18460605)
Son Lam Phung (18460602)
Soan T. M. Duong (18460599)
Abdesselam Bouzerdoum (17900021)
dc.date.none.fl_str_mv 2023-08-11T06:00:00Z
dc.identifier.none.fl_str_mv 10.1007/s10489-023-04682-6
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/MSD-NAS_multi-scale_dense_neural_architecture_search_for_real-time_pedestrian_lane_detection/26788183
dc.rights.none.fl_str_mv CC BY 4.0
info:eu-repo/semantics/openAccess
dc.subject.none.fl_str_mv Information and computing sciences
Artificial intelligence
Machine learning
Pedestrian lane detection
Real-time video processing
Neural architecture search
Assistive navigation
Deep learning
Semantic segmentation
dc.title.none.fl_str_mv MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p dir="ltr">Accurate detection of pedestrian lanes is a crucial criterion for vision-impaired people to navigate freely and safely. The current deep learning methods have achieved reasonable accuracy at this task. However, they lack practicality for real-time pedestrian lane detection due to non-optimal accuracy, speed, and model size trade-off. Hence, an optimized deep neural network (DNN) for pedestrian lane detection is required. Designing a DNN from scratch is a laborious task that requires significant experience and time. This paper proposes a novel neural architecture search (NAS) algorithm, named MSD-NAS, to automate this laborious task. The proposed method designs an optimized deep network with multi-scale input branches, allowing the derived network to utilize local and global contexts for predictions. The search is also performed in a large and generic space that includes many existing hand-designed network architectures as candidates. To further boost performance, we propose a Short-term Visual Memory mechanism to improve information facilitation within the derived networks. Evaluated on the PLVP3 dataset of 10,000 images, the DNN designed by MSD-NAS achieves state-of-the-art accuracy (0.9781) and mIoU (0.9542), while being 20.16 times faster and 2.56 times smaller than the current best deep learning model.</p><h2>Other Information</h2><p dir="ltr">Published in: Applied Intelligence<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1007/s10489-023-04682-6" target="_blank">https://dx.doi.org/10.1007/s10489-023-04682-6</a></p>
eu_rights_str_mv openAccess
id Manara2_0718f31f3281bd57cb611769082a36cc
identifier_str_mv 10.1007/s10489-023-04682-6
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/26788183
publishDate 2023
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detectionSui Paul Ang (18460605)Son Lam Phung (18460602)Soan T. M. Duong (18460599)Abdesselam Bouzerdoum (17900021)Information and computing sciencesArtificial intelligenceMachine learningPedestrian lane detectionReal-time video processingNeural architecture searchAssistive navigationDeep learningSemantic segmentation<p dir="ltr">Accurate detection of pedestrian lanes is a crucial criterion for vision-impaired people to navigate freely and safely. The current deep learning methods have achieved reasonable accuracy at this task. However, they lack practicality for real-time pedestrian lane detection due to non-optimal accuracy, speed, and model size trade-off. Hence, an optimized deep neural network (DNN) for pedestrian lane detection is required. Designing a DNN from scratch is a laborious task that requires significant experience and time. This paper proposes a novel neural architecture search (NAS) algorithm, named MSD-NAS, to automate this laborious task. The proposed method designs an optimized deep network with multi-scale input branches, allowing the derived network to utilize local and global contexts for predictions. The search is also performed in a large and generic space that includes many existing hand-designed network architectures as candidates. To further boost performance, we propose a Short-term Visual Memory mechanism to improve information facilitation within the derived networks. Evaluated on the PLVP3 dataset of 10,000 images, the DNN designed by MSD-NAS achieves state-of-the-art accuracy (0.9781) and mIoU (0.9542), while being 20.16 times faster and 2.56 times smaller than the current best deep learning model.</p><h2>Other Information</h2><p dir="ltr">Published in: Applied Intelligence<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1007/s10489-023-04682-6" target="_blank">https://dx.doi.org/10.1007/s10489-023-04682-6</a></p>2023-08-11T06:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1007/s10489-023-04682-6https://figshare.com/articles/journal_contribution/MSD-NAS_multi-scale_dense_neural_architecture_search_for_real-time_pedestrian_lane_detection/26788183CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/267881832023-08-11T06:00:00Z
spellingShingle MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
Sui Paul Ang (18460605)
Information and computing sciences
Artificial intelligence
Machine learning
Pedestrian lane detection
Real-time video processing
Neural architecture search
Assistive navigation
Deep learning
Semantic segmentation
status_str publishedVersion
title MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
title_full MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
title_fullStr MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
title_full_unstemmed MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
title_short MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
title_sort MSD-NAS: multi-scale dense neural architecture search for real-time pedestrian lane detection
topic Information and computing sciences
Artificial intelligence
Machine learning
Pedestrian lane detection
Real-time video processing
Neural architecture search
Assistive navigation
Deep learning
Semantic segmentation