A Dynamic Adjust‐Head Siamese network for object tracking
Abstract Siamese network based trackers formulate tracking as a similarity matching problem between a target template and a search region. Virtually all popular Siamese trackers use cross‐correlation to measure the similarity between the deep feature of template and search image. However, the emphas...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2023-03-01
|
| Series: | IET Computer Vision |
| Online Access: | https://doi.org/10.1049/cvi2.12148 |
| _version_ | 1797854645467480064 |
|---|---|
| author | Shoumeng Qiu Yuzhang Gu Minghong Chen Zeqiang Yuan Zehao Yao Xiaolin Zhang |
| author_facet | Shoumeng Qiu Yuzhang Gu Minghong Chen Zeqiang Yuan Zehao Yao Xiaolin Zhang |
| author_sort | Shoumeng Qiu |
| collection | DOAJ |
| description | Abstract Siamese network based trackers formulate tracking as a similarity matching problem between a target template and a search region. Virtually all popular Siamese trackers use cross‐correlation to measure the similarity between the deep feature of template and search image. However, the emphasis for feature extraction in different parts of the image are the same. Besides, the global matching between the template and search region also seriously neglects the part‐level information and the deformation of targets during tracking. In this study, to tackle the above issues, a simple but effective Dynamic Adjust‐Head (SiamDAH) model is proposed to extract features from different parts of an object. In addition, an improved pixelwise cross‐correlation model (PWCC) is designed to enhance the naive cross‐correlation operation to produce multiple similarity maps associated with different parts of the target. Experiments on serval challenging benchmarks including OTB‐100, GOT‐10k, LaSOT, and TrackingNet demonstrate that the proposed SiamDAH outperforms many state‐of‐the‐art trackers and achieves leading performance. |
| first_indexed | 2024-04-09T20:10:32Z |
| format | Article |
| id | doaj.art-13844f4556ad4bffb7a0a78de49db9d3 |
| institution | Directory Open Access Journal |
| issn | 1751-9632 1751-9640 |
| language | English |
| last_indexed | 2024-04-09T20:10:32Z |
| publishDate | 2023-03-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Computer Vision |
| spelling | doaj.art-13844f4556ad4bffb7a0a78de49db9d32023-04-01T03:37:25ZengWileyIET Computer Vision1751-96321751-96402023-03-0117220321010.1049/cvi2.12148A Dynamic Adjust‐Head Siamese network for object trackingShoumeng Qiu0Yuzhang Gu1Minghong Chen2Zeqiang Yuan3Zehao Yao4Xiaolin Zhang5Bio‐Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai ChinaBio‐Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai ChinaBio‐Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai ChinaBio‐Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai ChinaBio‐Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai ChinaBio‐Vision System Laboratory State Key Laboratory of Transducer Technology Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai ChinaAbstract Siamese network based trackers formulate tracking as a similarity matching problem between a target template and a search region. Virtually all popular Siamese trackers use cross‐correlation to measure the similarity between the deep feature of template and search image. However, the emphasis for feature extraction in different parts of the image are the same. Besides, the global matching between the template and search region also seriously neglects the part‐level information and the deformation of targets during tracking. In this study, to tackle the above issues, a simple but effective Dynamic Adjust‐Head (SiamDAH) model is proposed to extract features from different parts of an object. In addition, an improved pixelwise cross‐correlation model (PWCC) is designed to enhance the naive cross‐correlation operation to produce multiple similarity maps associated with different parts of the target. Experiments on serval challenging benchmarks including OTB‐100, GOT‐10k, LaSOT, and TrackingNet demonstrate that the proposed SiamDAH outperforms many state‐of‐the‐art trackers and achieves leading performance.https://doi.org/10.1049/cvi2.12148 |
| spellingShingle | Shoumeng Qiu Yuzhang Gu Minghong Chen Zeqiang Yuan Zehao Yao Xiaolin Zhang A Dynamic Adjust‐Head Siamese network for object tracking IET Computer Vision |
| title | A Dynamic Adjust‐Head Siamese network for object tracking |
| title_full | A Dynamic Adjust‐Head Siamese network for object tracking |
| title_fullStr | A Dynamic Adjust‐Head Siamese network for object tracking |
| title_full_unstemmed | A Dynamic Adjust‐Head Siamese network for object tracking |
| title_short | A Dynamic Adjust‐Head Siamese network for object tracking |
| title_sort | dynamic adjust head siamese network for object tracking |
| url | https://doi.org/10.1049/cvi2.12148 |
| work_keys_str_mv | AT shoumengqiu adynamicadjustheadsiamesenetworkforobjecttracking AT yuzhanggu adynamicadjustheadsiamesenetworkforobjecttracking AT minghongchen adynamicadjustheadsiamesenetworkforobjecttracking AT zeqiangyuan adynamicadjustheadsiamesenetworkforobjecttracking AT zehaoyao adynamicadjustheadsiamesenetworkforobjecttracking AT xiaolinzhang adynamicadjustheadsiamesenetworkforobjecttracking AT shoumengqiu dynamicadjustheadsiamesenetworkforobjecttracking AT yuzhanggu dynamicadjustheadsiamesenetworkforobjecttracking AT minghongchen dynamicadjustheadsiamesenetworkforobjecttracking AT zeqiangyuan dynamicadjustheadsiamesenetworkforobjecttracking AT zehaoyao dynamicadjustheadsiamesenetworkforobjecttracking AT xiaolinzhang dynamicadjustheadsiamesenetworkforobjecttracking |