Siamese Transformer Network for Real-Time Aerial Object Tracking
Recently, deep learning (DL) based trackers have attracted tremendous interest for their high performance. Despite the remarkable success, most trackers utilizing deep convolution features commonly neglect tracking speed, which is crucial for aerial tracking on mobile devices. In this paper, we prop...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9908547/ |
| _version_ | 1817986242338357248 |
|---|---|
| author | Haijun Wang Shengyan Zhang |
| author_facet | Haijun Wang Shengyan Zhang |
| author_sort | Haijun Wang |
| collection | DOAJ |
| description | Recently, deep learning (DL) based trackers have attracted tremendous interest for their high performance. Despite the remarkable success, most trackers utilizing deep convolution features commonly neglect tracking speed, which is crucial for aerial tracking on mobile devices. In this paper, we propose an efficient and effective transformer based aerial tracker in the framework of Siamese, which inherits the merits from both transformer and Siamese architectures. Specifically, the outputs from multiple convolution layers are fed into transformer to construct robust features of template patch and search patch, respectively. Consequently, the interdependencies between low-level information and semantic information are interactively fused to improve the ability of encoding target appearance. Finally, traditional depth-wise cross correlation is introduced to generate a similarity map for object location and bounding box regression. Extensive experimental results on three popular benchmarks (DTB70, UAV123@10fps, and UAV20L) have demonstrated that our proposed tracker outperforms other 12 state-of-the-art trackers and achieves a real-time tracking speed of 71.3 frames per second (FPS) on GPU, which can be applied in mobile platform. |
| first_indexed | 2024-04-14T00:07:51Z |
| format | Article |
| id | doaj.art-862fe681a48d4e018513c31d7ce49dab |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-14T00:07:51Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-862fe681a48d4e018513c31d7ce49dab2022-12-22T02:23:28ZengIEEEIEEE Access2169-35362022-01-011010520110521310.1109/ACCESS.2022.32115169908547Siamese Transformer Network for Real-Time Aerial Object TrackingHaijun Wang0https://orcid.org/0000-0003-2481-9662Shengyan Zhang1Aviation Information Technology Research and Development, Binzhou University, Binzhou, ChinaAviation Information Technology Research and Development, Binzhou University, Binzhou, ChinaRecently, deep learning (DL) based trackers have attracted tremendous interest for their high performance. Despite the remarkable success, most trackers utilizing deep convolution features commonly neglect tracking speed, which is crucial for aerial tracking on mobile devices. In this paper, we propose an efficient and effective transformer based aerial tracker in the framework of Siamese, which inherits the merits from both transformer and Siamese architectures. Specifically, the outputs from multiple convolution layers are fed into transformer to construct robust features of template patch and search patch, respectively. Consequently, the interdependencies between low-level information and semantic information are interactively fused to improve the ability of encoding target appearance. Finally, traditional depth-wise cross correlation is introduced to generate a similarity map for object location and bounding box regression. Extensive experimental results on three popular benchmarks (DTB70, UAV123@10fps, and UAV20L) have demonstrated that our proposed tracker outperforms other 12 state-of-the-art trackers and achieves a real-time tracking speed of 71.3 frames per second (FPS) on GPU, which can be applied in mobile platform.https://ieeexplore.ieee.org/document/9908547/Aerial object trackingtransformerself-attentionSiamese networkdeep neural network |
| spellingShingle | Haijun Wang Shengyan Zhang Siamese Transformer Network for Real-Time Aerial Object Tracking IEEE Access Aerial object tracking transformer self-attention Siamese network deep neural network |
| title | Siamese Transformer Network for Real-Time Aerial Object Tracking |
| title_full | Siamese Transformer Network for Real-Time Aerial Object Tracking |
| title_fullStr | Siamese Transformer Network for Real-Time Aerial Object Tracking |
| title_full_unstemmed | Siamese Transformer Network for Real-Time Aerial Object Tracking |
| title_short | Siamese Transformer Network for Real-Time Aerial Object Tracking |
| title_sort | siamese transformer network for real time aerial object tracking |
| topic | Aerial object tracking transformer self-attention Siamese network deep neural network |
| url | https://ieeexplore.ieee.org/document/9908547/ |
| work_keys_str_mv | AT haijunwang siamesetransformernetworkforrealtimeaerialobjecttracking AT shengyanzhang siamesetransformernetworkforrealtimeaerialobjecttracking |