Market1501TODukeMTMC loss ce.
<div><p>Person re-identification (ReID) technology has many applications in intelligent surveillance and public safety. However, the domain difference between the source and target domains makes the generalization ability of the model extremely challenging. To reduce the dependence on la...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Published: |
2025
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1852018491570257920 |
|---|---|
| author | Xuemei Bai (169120) |
| author2 | Yuqing Zhang (32031) Chenjie Zhang (11328211) Zhijun Wang (31218) |
| author2_role | author author author |
| author_facet | Xuemei Bai (169120) Yuqing Zhang (32031) Chenjie Zhang (11328211) Zhijun Wang (31218) |
| author_role | author |
| dc.creator.none.fl_str_mv | Xuemei Bai (169120) Yuqing Zhang (32031) Chenjie Zhang (11328211) Zhijun Wang (31218) |
| dc.date.none.fl_str_mv | 2025-07-14T17:36:15Z |
| dc.identifier.none.fl_str_mv | 10.1371/journal.pone.0328131.g005 |
| dc.relation.none.fl_str_mv | https://figshare.com/articles/figure/Market1501TODukeMTMC_loss_ce_/29564211 |
| dc.rights.none.fl_str_mv | CC BY 4.0 info:eu-repo/semantics/openAccess |
| dc.subject.none.fl_str_mv | Neuroscience Space Science Biological Sciences not elsewhere classified mean average precision innovative loss function target domains makes scale public datasets effective technical support domain migration performance 9 %, respectively model extremely challenging limited model performance unsupervised domain adaptation label noise processing label generated noise xlink "> person contrast learning technique unsupervised domain adapted identification </ p existing uda methods unsupervised domain label noise target domain contrast learning public safety label refinement effective way domain difference model training two large training process thereby enhancing still significant probabilistic uncertainty paper provides paper proposes negative impact method experiments many applications labeled data intelligent surveillance identification task good solution generation process generalization ability first enhance feature space feature representation experimental validation 4 %, 1 accuracy |
| dc.title.none.fl_str_mv | Market1501TODukeMTMC loss ce. |
| dc.type.none.fl_str_mv | Image Figure info:eu-repo/semantics/publishedVersion image |
| description | <div><p>Person re-identification (ReID) technology has many applications in intelligent surveillance and public safety. However, the domain difference between the source and target domains makes the generalization ability of the model extremely challenging. To reduce the dependence on labeled data, Unsupervised Domain Adaptation (UDA) methods have become an effective way to solve this problem. However, the influence of pseudo-label generated noise on model training in existing UDA methods is still significant, resulting in limited model performance on the target domain. For this reason, this paper proposes a contrast learning-based pseudo-label refinement with probabilistic uncertainty in the unsupervised domain, adapted to Person re-identification, aiming to improve the effectiveness of the unsupervised domain adapted to Person re-identification. We first enhance the feature representation of the target domain samples based on the contrast learning technique to improve their discrimination in the feature space, thereby enhancing the cross-domain migration performance of the model. Subsequently, an innovative loss function is proposed to effectively reduce the interference of label noise on the training process by refining the generation process of pseudo-labels, which solves the negative impact of inaccurate pseudo-labels on model training. Through a series of experimental validation, the method experiments on two large-scale public datasets, Market1501 and DukeMTMC, and the Rank-1 accuracy of the proposed method reaches 91.4% and 81.4%, with the mean average precision (mAP) of 79.0% and 67.9%, respectively, which proves that the research in this paper provides a good solution for the Person re-identification task with effective technical support for label noise processing and model generalization capability improvement.</p></div> |
| eu_rights_str_mv | openAccess |
| id | Manara_fc2ad2a9c33ecb7244e511fe899a2ae8 |
| identifier_str_mv | 10.1371/journal.pone.0328131.g005 |
| network_acronym_str | Manara |
| network_name_str | ManaraRepo |
| oai_identifier_str | oai:figshare.com:article/29564211 |
| publishDate | 2025 |
| repository.mail.fl_str_mv | |
| repository.name.fl_str_mv | |
| repository_id_str | |
| rights_invalid_str_mv | CC BY 4.0 |
| spelling | Market1501TODukeMTMC loss ce.Xuemei Bai (169120)Yuqing Zhang (32031)Chenjie Zhang (11328211)Zhijun Wang (31218)NeuroscienceSpace ScienceBiological Sciences not elsewhere classifiedmean average precisioninnovative loss functiontarget domains makesscale public datasetseffective technical supportdomain migration performance9 %, respectivelymodel extremely challenginglimited model performanceunsupervised domain adaptationlabel noise processinglabel generated noisexlink "> personcontrast learning techniqueunsupervised domain adaptedidentification </ pexisting uda methodsunsupervised domainlabel noisetarget domaincontrast learningpublic safetylabel refinementeffective waydomain differencemodel trainingtwo largetraining processthereby enhancingstill significantprobabilistic uncertaintypaper providespaper proposesnegative impactmethod experimentsmany applicationslabeled dataintelligent surveillanceidentification taskgood solutiongeneration processgeneralization abilityfirst enhancefeature spacefeature representationexperimental validation4 %,1 accuracy<div><p>Person re-identification (ReID) technology has many applications in intelligent surveillance and public safety. However, the domain difference between the source and target domains makes the generalization ability of the model extremely challenging. To reduce the dependence on labeled data, Unsupervised Domain Adaptation (UDA) methods have become an effective way to solve this problem. However, the influence of pseudo-label generated noise on model training in existing UDA methods is still significant, resulting in limited model performance on the target domain. For this reason, this paper proposes a contrast learning-based pseudo-label refinement with probabilistic uncertainty in the unsupervised domain, adapted to Person re-identification, aiming to improve the effectiveness of the unsupervised domain adapted to Person re-identification. We first enhance the feature representation of the target domain samples based on the contrast learning technique to improve their discrimination in the feature space, thereby enhancing the cross-domain migration performance of the model. Subsequently, an innovative loss function is proposed to effectively reduce the interference of label noise on the training process by refining the generation process of pseudo-labels, which solves the negative impact of inaccurate pseudo-labels on model training. Through a series of experimental validation, the method experiments on two large-scale public datasets, Market1501 and DukeMTMC, and the Rank-1 accuracy of the proposed method reaches 91.4% and 81.4%, with the mean average precision (mAP) of 79.0% and 67.9%, respectively, which proves that the research in this paper provides a good solution for the Person re-identification task with effective technical support for label noise processing and model generalization capability improvement.</p></div>2025-07-14T17:36:15ZImageFigureinfo:eu-repo/semantics/publishedVersionimage10.1371/journal.pone.0328131.g005https://figshare.com/articles/figure/Market1501TODukeMTMC_loss_ce_/29564211CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/295642112025-07-14T17:36:15Z |
| spellingShingle | Market1501TODukeMTMC loss ce. Xuemei Bai (169120) Neuroscience Space Science Biological Sciences not elsewhere classified mean average precision innovative loss function target domains makes scale public datasets effective technical support domain migration performance 9 %, respectively model extremely challenging limited model performance unsupervised domain adaptation label noise processing label generated noise xlink "> person contrast learning technique unsupervised domain adapted identification </ p existing uda methods unsupervised domain label noise target domain contrast learning public safety label refinement effective way domain difference model training two large training process thereby enhancing still significant probabilistic uncertainty paper provides paper proposes negative impact method experiments many applications labeled data intelligent surveillance identification task good solution generation process generalization ability first enhance feature space feature representation experimental validation 4 %, 1 accuracy |
| status_str | publishedVersion |
| title | Market1501TODukeMTMC loss ce. |
| title_full | Market1501TODukeMTMC loss ce. |
| title_fullStr | Market1501TODukeMTMC loss ce. |
| title_full_unstemmed | Market1501TODukeMTMC loss ce. |
| title_short | Market1501TODukeMTMC loss ce. |
| title_sort | Market1501TODukeMTMC loss ce. |
| topic | Neuroscience Space Science Biological Sciences not elsewhere classified mean average precision innovative loss function target domains makes scale public datasets effective technical support domain migration performance 9 %, respectively model extremely challenging limited model performance unsupervised domain adaptation label noise processing label generated noise xlink "> person contrast learning technique unsupervised domain adapted identification </ p existing uda methods unsupervised domain label noise target domain contrast learning public safety label refinement effective way domain difference model training two large training process thereby enhancing still significant probabilistic uncertainty paper provides paper proposes negative impact method experiments many applications labeled data intelligent surveillance identification task good solution generation process generalization ability first enhance feature space feature representation experimental validation 4 %, 1 accuracy |