Dual Attentional Siamese Network for Visual Tracking

Zhang, Xiaowei; Ma, Jianwei; Liu, Hong; Hu, Hai‐Miao; Yang, Peng

doi:10.1016/j.displa.2022.102205

Cited by 5 publications

(2 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They combined this with spatial and channel attention mechanisms, achieving improved scene segmentation results by capturing long-range contextual dependencies. Furthermore, Zhang et al [25] in their Dual Attentional Siamese Network (DASNet) utilized dilated convolutions in a unique framework for visual tracking. The architecture adeptly captures fine-grained and global semantic consistencies, again demonstrating the utility of dilated convolutions in challenging vision tasks.…”

Section: Dilated Convolutionmentioning

confidence: 99%

Hyper-Progressive Single Shot Detector (HPSSD) Algorithm for Door Panel Type-B Detection

Yuan,

Okpor,

Kong

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In an automobile, the door panel type-B constitutes the interior compartment of the door, primarily composed of screws and white installations forming its structural framework. However, automated manufacturing and maintenance procedures often struggle to accurately detect these components due to their pronounced resemblance to other elements on the panel. Computer vision techniques present a viable solution to this challenge. In this paper, we propose the Hyper-Progressive Single Shot Detector (HPSSD), an object detection algorithm designed to address the aforementioned challenge. Our proposed HPSSD builds on the Single Shot Detector (SSD) algorithm and introduces several enhancements to improve its detection capabilities. The first modification involves replacing the VGG-16 backbone with a ResNet-50 module. Furthermore, we incorporated the Residual Convolutional Block Attention Mechanism (RCBAM) to boost the algorithm's functionality. To enlarge the receptive fields of each pixel-an essential step for enhancing detection accuracy-we executed multi-dilated convolutions. In the final stage of our development process, we embedded a three-stage progressive attention mechanism (PAM). The PAM is instrumental in generating refined feature maps, which serve as the foundation for precise object detection on the door panel dataset comprising 1200 images. After running 50k iterations on the door panel dataset, the HPSSD displayed a promising mean average precision of 98.2% at a speed of 21 frames per second (FPS). Our results suggest that the HPSSD, with its ability to deliver real-time, accurate detection, is an ideal tool for improving the quality inspection of door panels in smart factories. INDEX TERMSDoor panel inspection, Hyper-progressive single shot detector (HPSSD), Multi-dilated convolution module (MDCM), Object detection, Progressive attention module (PAM), Residual convolutional block attention mechanism (RCBAM), Smart factories.

show abstract

Section: Dilated Convolutionmentioning

confidence: 99%

Hyper-Progressive Single Shot Detector (HPSSD) Algorithm for Door Panel Type-B Detection

Yuan,

Okpor,

Kong

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…To handle various challenges, a significant amount of research has been focused on visual tracking in recent years. With the continuous advancement of machine learning and signal processing technology, algorithms based on correlation filters (Zhang J. et al, 2022 ; Ly et al, 2021 ) and deep learning (Haisheng et al, 2019 ; Voigtlaender et al, 2020 ; Tan et al, 2021 ; Zhang X. et al, 2022 ) have gradually replaced traditional methods as the mainstream object tracking algorithms. Both of these methods are discriminative methods.…”

Section: Introductionmentioning

confidence: 99%

A Siamese tracker with “dynamic–static” dual-template fusion and dynamic template adaptive update

et al. 2023

View full text Add to dashboard Cite

In recent years, visual tracking algorithms based on Siamese networks have attracted attention for their desirable balance between speed and accuracy. The performance of such tracking methods relies heavily on target templates. Static templates cannot cope with the adverse effects of target appearance change. The dynamic template method, with a template update mechanism, can adapt to the change in target appearance well, but it also causes new problems, which may lead the template to be polluted by noise. Based on the DaSiamRPN and UpdateNet template update networks, a Siamese tracker with “dynamic–static” dual-template fusion and dynamic template adaptive update is proposed in this paper. The new method combines a static template and a dynamic template that is updated in real time for object tracking. An adaptive update strategy was adopted when updating the dynamic template, which can not only help adjust to the changes in the object appearance, but also suppress the adverse effects of noise interference and contamination of the template. The experimental results showed that the robustness and EAO of the proposed method were 23% and 9.0% higher than those of the basic algorithm on the VOT2016 dataset, respectively, and that the precision and success were increased by 0.8 and 0.4% on the OTB100 dataset, respectively. The most comprehensive real-time tracking performance was obtained for the above two large public datasets.

show abstract