POSNet: a hybrid deep learning model for efficient person re-identification
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Batool, E. | - |
dc.contributor.author | Gillani, S. | - |
dc.contributor.author | Naz, S. | - |
dc.contributor.author | Bukhari, M. | - |
dc.contributor.author | Maqsood, M. | - |
dc.contributor.author | Yeo, S.-S. | - |
dc.contributor.author | Rho, Seungmin | - |
dc.date.accessioned | 2024-01-09T06:06:40Z | - |
dc.date.available | 2024-01-09T06:06:40Z | - |
dc.date.issued | 2023-08 | - |
dc.identifier.issn | 0920-8542 | - |
dc.identifier.issn | 1573-0484 | - |
dc.identifier.uri | https://scholarworks.bwise.kr/cau/handle/2019.sw.cau/69979 | - |
dc.description.abstract | Person re-identification refers to the process of recognizing a person across several non-overlapping cameras. It is becoming increasingly important in computer vision for real-world surveillance applications. However, the deployment of person-re-identification systems as a surveillance system raises various challenges in their performance. These challenges include limited labeled data, occlusions conditions, human body postures, as well as inter- and intra-class variations. Such challenges deteriorate the effectiveness of person-re-identification systems and lead to the extraction of less discriminative features. Hence, to address these problems, we proposed a hybrid deep learning model, namely POSNet (pseudo-labeled omni-scale network) for efficient person re-identification. The proposed method is referred to as a hybrid because it combines label estimate with modified omni-scale feature learning, i.e., spatiotemporal-assisted omni-scale feature extraction to accomplish person re-identification. To further enhance omni-scale feature learning, we have proposed soft-pool-assisted attention mechanisms during spatial learning. More precisely, soft-pool preserves more important features, and that features are further emphasized by spatial and channel attention layers. Following on, this omni-scale with soft-pool attention learning extracts the spatial information from all frames of videos, and later on, the temporal learning is incorporated using the LSTM model. To handle limited labeled data problems, the proposed hybrid model first assigns pseudo-labels to the unlabeled data and adopts a progressive learning strategy to retrain the model on both labeled and unlabeled data with improved feature extraction, i.e., modified omni-scale feature learning. Moreover, the proposed POSNet model is validated on two large video-based person re-identification datasets, namely MARS and DukeMTMC-Video. It is observed from the research findings that the proposed POSNet outperformed the existing studies with the highest mAP and rank@1 score of 83.7 and 90.3%, respectively. © 2023, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature. | - |
dc.format.extent | 29 | - |
dc.language | 영어 | - |
dc.language.iso | ENG | - |
dc.publisher | Springer | - |
dc.title | POSNet: a hybrid deep learning model for efficient person re-identification | - |
dc.type | Article | - |
dc.identifier.doi | 10.1007/s11227-023-05169-4 | - |
dc.identifier.bibliographicCitation | Journal of Supercomputing, v.79, no.12, pp 13090 - 13118 | - |
dc.description.isOpenAccess | N | - |
dc.identifier.wosid | 000954213300001 | - |
dc.identifier.scopusid | 2-s2.0-85150424863 | - |
dc.citation.endPage | 13118 | - |
dc.citation.number | 12 | - |
dc.citation.startPage | 13090 | - |
dc.citation.title | Journal of Supercomputing | - |
dc.citation.volume | 79 | - |
dc.type.docType | Article | - |
dc.publisher.location | 네델란드 | - |
dc.subject.keywordAuthor | Hybrid deep learning | - |
dc.subject.keywordAuthor | Intra- and inter-class variations | - |
dc.subject.keywordAuthor | Limited labeled data challenges | - |
dc.subject.keywordAuthor | LSTM | - |
dc.subject.keywordAuthor | Person re-identification | - |
dc.subject.keywordAuthor | POSNet | - |
dc.subject.keywordAuthor | Soft-pool-assisted attentions | - |
dc.subject.keywordAuthor | Spatio-temporal feature learning | - |
dc.relation.journalResearchArea | Computer Science | - |
dc.relation.journalResearchArea | Engineering | - |
dc.relation.journalWebOfScienceCategory | Computer Science, Hardware & Architecture | - |
dc.relation.journalWebOfScienceCategory | Computer Science, Theory & Methods | - |
dc.relation.journalWebOfScienceCategory | Engineering, Electrical & Electronic | - |
dc.description.journalRegisteredClass | scie | - |
dc.description.journalRegisteredClass | scopus | - |
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.
84, Heukseok-ro, Dongjak-gu, Seoul, Republic of Korea (06974)02-820-6194
COPYRIGHT 2019 Chung-Ang University All Rights Reserved.
Certain data included herein are derived from the © Web of Science of Clarivate Analytics. All rights reserved.
You may not copy or re-distribute this material in whole or in part without the prior written consent of Clarivate Analytics.