Online Siamese Network for Visual Object Tracking
Offline-trained Siamese networks are not robust to the environmental complication in visual object tracking. Without online learning, the Siamese network cannot learn from instance domain knowledge and adapt to appearance changes of targets. In this paper, a new lightweight Siamese network is propos...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2019-04-01
|
Series: | Sensors |
Subjects: | |
Online Access: | https://www.mdpi.com/1424-8220/19/8/1858 |
id |
doaj-9020e24924f24bf0912e46519a7158af |
---|---|
record_format |
Article |
spelling |
doaj-9020e24924f24bf0912e46519a7158af2020-11-25T00:58:53ZengMDPI AGSensors1424-82202019-04-01198185810.3390/s19081858s19081858Online Siamese Network for Visual Object TrackingShuo Chang0Wei Li1Yifan Zhang2Zhiyong Feng3School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaDepartment of Electrical Engineering, Northern Illinois University, Dekalb, IL 60115, USASchool of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaSchool of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaOffline-trained Siamese networks are not robust to the environmental complication in visual object tracking. Without online learning, the Siamese network cannot learn from instance domain knowledge and adapt to appearance changes of targets. In this paper, a new lightweight Siamese network is proposed for feature extraction. To cope with the dynamics of targets and backgrounds, the weight in the proposed Siamese network is updated in an online manner during the tracking process. In order to enhance the discrimination capability, the cross-entropy loss is integrated into the contrastive loss. Inspired by the face verification algorithm DeepID2, the Bayesian verification model is applied for candidate selection. In general, visual object tracking can benefit from face verification algorithms. Numerical results suggest that the newly developed algorithm achieves comparable performance in public benchmarks.https://www.mdpi.com/1424-8220/19/8/1858visual object trackingSiamese networkimproved contrastive lossBayesian verification |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Shuo Chang Wei Li Yifan Zhang Zhiyong Feng |
spellingShingle |
Shuo Chang Wei Li Yifan Zhang Zhiyong Feng Online Siamese Network for Visual Object Tracking Sensors visual object tracking Siamese network improved contrastive loss Bayesian verification |
author_facet |
Shuo Chang Wei Li Yifan Zhang Zhiyong Feng |
author_sort |
Shuo Chang |
title |
Online Siamese Network for Visual Object Tracking |
title_short |
Online Siamese Network for Visual Object Tracking |
title_full |
Online Siamese Network for Visual Object Tracking |
title_fullStr |
Online Siamese Network for Visual Object Tracking |
title_full_unstemmed |
Online Siamese Network for Visual Object Tracking |
title_sort |
online siamese network for visual object tracking |
publisher |
MDPI AG |
series |
Sensors |
issn |
1424-8220 |
publishDate |
2019-04-01 |
description |
Offline-trained Siamese networks are not robust to the environmental complication in visual object tracking. Without online learning, the Siamese network cannot learn from instance domain knowledge and adapt to appearance changes of targets. In this paper, a new lightweight Siamese network is proposed for feature extraction. To cope with the dynamics of targets and backgrounds, the weight in the proposed Siamese network is updated in an online manner during the tracking process. In order to enhance the discrimination capability, the cross-entropy loss is integrated into the contrastive loss. Inspired by the face verification algorithm DeepID2, the Bayesian verification model is applied for candidate selection. In general, visual object tracking can benefit from face verification algorithms. Numerical results suggest that the newly developed algorithm achieves comparable performance in public benchmarks. |
topic |
visual object tracking Siamese network improved contrastive loss Bayesian verification |
url |
https://www.mdpi.com/1424-8220/19/8/1858 |
work_keys_str_mv |
AT shuochang onlinesiamesenetworkforvisualobjecttracking AT weili onlinesiamesenetworkforvisualobjecttracking AT yifanzhang onlinesiamesenetworkforvisualobjecttracking AT zhiyongfeng onlinesiamesenetworkforvisualobjecttracking |
_version_ |
1725220103725776896 |