DSpace Repository

Target-Specific Siamese Attention Network for Real-time Object Tracking

Show simple item record

dc.contributor.author Kokul, T.
dc.contributor.author Fookes, C.
dc.contributor.author Sridharan, S.
dc.contributor.author Ramanan, A.
dc.contributor.author Pinidiyaarachchi, A.
dc.date.accessioned 2020-01-08T05:19:34Z
dc.date.accessioned 2022-06-27T04:11:18Z
dc.date.available 2020-01-08T05:19:34Z
dc.date.available 2022-06-27T04:11:18Z
dc.date.issued 2019-08-16
dc.identifier.issn 1556-6021
dc.identifier.uri http://repo.lib.jfn.ac.lk/ujrr/handle/123456789/1312
dc.description.abstract Deep similarity trackers are able to track above real-time speed. However, their accuracy is considerably lower than deep classification based trackers since they avoid valuable online cues. To feed the target-specific information for real-time object tracking, we propose a novel Siamese attention network. Different types of attention mechanisms are used to capture different contexts of target information and then learned knowledge is used to feed target cues at different representation levels of similarity tracking. In addition, an online learning mechanism is employed to utilise the available target-specific data. The proposed tracker reduces the impact of noise in the target template and improves the accuracy of similarity tracking by feeding target cues into the similarity search. Extensive evaluation performed on OTB-2013/50/100 and VOT2018 benchmark datasets demonstrate the proposed tracker outperforms state-of-the-art approaches while maintaining real-time tracking speed. en_US
dc.language.iso en en_US
dc.publisher IEEE Transactions on Information Forensics and Security en_US
dc.subject Visual object tracking en_US
dc.subject Deep neural networks en_US
dc.subject Siamese network en_US
dc.subject Attention network en_US
dc.title Target-Specific Siamese Attention Network for Real-time Object Tracking en_US
dc.type Article en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record