End‐to‐end feature fusion Siamese network for adaptive visual tracking
Abstract According to observations, different visual objects have different salient features in different scenarios. Even for the same object, its salient shape and appearance features may change greatly from time to time in a long‐term tracking task. Motivated by them, an end‐to‐end feature fusion...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2021-01-01
|
Series: | IET Image Processing |
Online Access: | https://doi.org/10.1049/ipr2.12009 |
id |
doaj-39f951463ceb43d09637caf32807d831 |
---|---|
record_format |
Article |
spelling |
doaj-39f951463ceb43d09637caf32807d8312021-07-14T13:25:38ZengWileyIET Image Processing1751-96591751-96672021-01-011519110010.1049/ipr2.12009End‐to‐end feature fusion Siamese network for adaptive visual trackingDongyan Guo0Jun Wang1Weixuan Zhao2Ying Cui3Zhenhua Wang4Shengyong Chen5College of Computer Science and Technology Zhejiang University of Technology Hangzhou Zhejiang ChinaCollege of Computer Science and Technology Zhejiang University of Technology Hangzhou Zhejiang ChinaCollege of Computer Science and Technology Zhejiang University of Technology Hangzhou Zhejiang ChinaCollege of Computer Science and Technology Zhejiang University of Technology Hangzhou Zhejiang ChinaCollege of Computer Science and Technology Zhejiang University of Technology Hangzhou Zhejiang ChinaSchool of Computer Science and Engineering Tianjin University of Technology Tianjin ChinaAbstract According to observations, different visual objects have different salient features in different scenarios. Even for the same object, its salient shape and appearance features may change greatly from time to time in a long‐term tracking task. Motivated by them, an end‐to‐end feature fusion framework was proposed based on the Siamese network, named FF‐Siam, which can effectively fuse different features for adaptive visual tracking. The framework consists of four layers. A feature extraction layer is designed to extract the different features of the target region and search region. The extracted features are then put into a weight generation layer to obtain the channel weights, which indicate the importance of different feature channels. Both features and the channel weights are utilised in a template generation layer to generate a discriminative template. Finally, the corresponding response maps created by the convolution of the search region features and the template are applied with a fusion layer to obtain the final response map for locating the target. Experimental results demonstrate that the proposed framework achieves state‐of‐the‐art performance on the popular Temple‐Colour, OTB50 and UAV123 benchmarks.https://doi.org/10.1049/ipr2.12009 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Dongyan Guo Jun Wang Weixuan Zhao Ying Cui Zhenhua Wang Shengyong Chen |
spellingShingle |
Dongyan Guo Jun Wang Weixuan Zhao Ying Cui Zhenhua Wang Shengyong Chen End‐to‐end feature fusion Siamese network for adaptive visual tracking IET Image Processing |
author_facet |
Dongyan Guo Jun Wang Weixuan Zhao Ying Cui Zhenhua Wang Shengyong Chen |
author_sort |
Dongyan Guo |
title |
End‐to‐end feature fusion Siamese network for adaptive visual tracking |
title_short |
End‐to‐end feature fusion Siamese network for adaptive visual tracking |
title_full |
End‐to‐end feature fusion Siamese network for adaptive visual tracking |
title_fullStr |
End‐to‐end feature fusion Siamese network for adaptive visual tracking |
title_full_unstemmed |
End‐to‐end feature fusion Siamese network for adaptive visual tracking |
title_sort |
end‐to‐end feature fusion siamese network for adaptive visual tracking |
publisher |
Wiley |
series |
IET Image Processing |
issn |
1751-9659 1751-9667 |
publishDate |
2021-01-01 |
description |
Abstract According to observations, different visual objects have different salient features in different scenarios. Even for the same object, its salient shape and appearance features may change greatly from time to time in a long‐term tracking task. Motivated by them, an end‐to‐end feature fusion framework was proposed based on the Siamese network, named FF‐Siam, which can effectively fuse different features for adaptive visual tracking. The framework consists of four layers. A feature extraction layer is designed to extract the different features of the target region and search region. The extracted features are then put into a weight generation layer to obtain the channel weights, which indicate the importance of different feature channels. Both features and the channel weights are utilised in a template generation layer to generate a discriminative template. Finally, the corresponding response maps created by the convolution of the search region features and the template are applied with a fusion layer to obtain the final response map for locating the target. Experimental results demonstrate that the proposed framework achieves state‐of‐the‐art performance on the popular Temple‐Colour, OTB50 and UAV123 benchmarks. |
url |
https://doi.org/10.1049/ipr2.12009 |
work_keys_str_mv |
AT dongyanguo endtoendfeaturefusionsiamesenetworkforadaptivevisualtracking AT junwang endtoendfeaturefusionsiamesenetworkforadaptivevisualtracking AT weixuanzhao endtoendfeaturefusionsiamesenetworkforadaptivevisualtracking AT yingcui endtoendfeaturefusionsiamesenetworkforadaptivevisualtracking AT zhenhuawang endtoendfeaturefusionsiamesenetworkforadaptivevisualtracking AT shengyongchen endtoendfeaturefusionsiamesenetworkforadaptivevisualtracking |
_version_ |
1721302675391250432 |