MediaWiki Research Articles

AwareTrack: Object awareness for visual tracking via templates interaction

Current popular trackers, whether based on the Siamese network or Transformer, have focused their main work on relation modeling between the template and the search area, and on the design of the tracking head, neglecting the fundamental element of tracking, the template. Templates are often mixed with too much background information, which can interfere with the extraction of template features. To address the above issue, a template object-aware tracker (AwareTrack) is proposed. Through the information interaction between multiple templates, the attention of the templates can be truly focused on the object itself, and the background interference can be suppressed. To ensure that the foreground objects of the templates have the same appearance to the greatest extent, the concept of awareness templates is proposed, which consists of two close frames. In addition, an awareness templates sampling method based on similarity discrimination via Siamese network is also proposed, which adaptively determines the interval between two awareness templates, ensure the maximization of background differences in the awareness templates. Meanwhile, online updates to the awareness templates ensure that our tracker has access to the most recent features of the foreground object. Our AwareTrack achieves state-of-the-art performance on multiple benchmarks, particularly on the one-shot tracking benchmark GOT-10k, achieving the AO of 78.1%, which is a 4.4% improvement over OSTrack-384.

About Wiki-To-Print

This brief statment explains the production of Content/Form from the designer' perspective: The journal is made with wiki-to-print, a collective publishing environment based on MediaWiki software, Paged Media CSS techniques and the JavaScript library Paged.js, which renders a preview of the PDF in the browser. Using wiki-to-print allows us to work shoulder-to-shoulder as collaborative writers, editors, designers, developers, in a non-linear publishing workflow where design and content unfolds at the same time, allowing the one to shape the other.

FETrack: Feature-Enhanced Transformer Network for Visual Object Tracking

Visual object tracking is a fundamental task in computer vision, with applications ranging from video surveillance to autonomous driving. Despite recent advances in transformer-based one-stream trackers, unrestricted feature interactions between the template and the search region often introduce background noise into the template, degrading the tracking performance. To address this issue, we propose FETrack, a feature-enhanced transformer-based network for visual object tracking. Specifically, we incorporate an independent template stream in the encoder of the one-stream tracker to acquire the high-quality template features while suppressing the harmful background noise effectively. Then, we employ a sequence-learning-based causal transformer in the decoder to generate the bounding box autoregressively, simplifying the prediction head network. Further, we present a dynamic threshold-based online template-updating strategy and a template-filtering approach to boost tracking robustness and reduce redundant computations. Extensive experiments demonstrate that our FETrack achieves a superior performance over state-of-the-art trackers. Specifically, the proposed FETrack achieves a 75.1% AO on GOT-10k, 81.2% AUC on LaSOT, and 89.3% Pnorm on TrackingNet.

Efficient transformer tracking with adaptive attention

AbstractRecently, several trackers utilising Transformer architecture have shown significant performance improvement. However, the high computational cost of multi‐head attention, a core component in the Transformer, has limited real‐time running speed, which is crucial for tracking tasks. Additionally, the global mechanism of multi‐head attention makes it susceptible to distractors with similar semantic information to the target. To address these issues, the authors propose a novel adaptive attention that enhances features through the spatial sparse attention mechanism with less than 1/4 of the computational complexity of multi‐head attention. Our adaptive attention sets a perception range around each element in the feature map based on the target scale in the previous tracking result and adaptively searches for the information of interest. This allows the module to focus on the target region rather than background distractors. Based on adaptive attention, the authors build an efficient transformer tracking framework. It can perform deep interaction between search and template features to activate target information and aggregate multi‐level interaction features to enhance the representation ability. The evaluation results on seven benchmarks show that the authors’ tracker achieves outstanding performance with a speed of 43 fps and significant advantages in hard circumstances.

CTIFTrack: Continuous Temporal Information Fusion for object track

In visual tracking tasks, researchers usually focus on increasing the complexity of the model or only discretely focusing on the changes in the object itself to achieve accurate recognition and tracking of the moving object. However, they often overlook the significant contribution of video-level linear temporal information fusion and continuous spatiotemporal mapping to tracking tasks. This oversight may lead to poor tracking performance or insufficient real-time ability of the model in complex scenes. Therefore, this paper proposes a real-time tracker, namely Continuous Temporal Information Fusion Tracker (CTIFTrack). The key of CTIFTrack lies in its well-designed Temporal Information Fusion (TIF) module, which cleverly performs a linear fusion of the temporal information between the (t-1)-th and the t-th frames and completes the spatiotemporal mapping. This enables the tracker to better understand the overall spatiotemporal information and contextual spatiotemporal correlations within the video, thereby having a positive impact on the tracking task. In addition, this paper also proposes the Object Template Feature Refinement (OTFR) module, which effectively captures the global information and local details of the object, and further improves the tracker’s understanding of the object features. Extensive experiments are conducted on seven benchmarks, such as LaSOT, GOT-10K, UAV123, NFS, TrackingNet, VOT2018 and OTB-100. The experimental results validate the significant contribution of the TIF module and OTFR module to the tracking task, as well as the effectiveness of CTIFTrack. It is worth noting that while maintaining excellent tracking performance, CTIFTrack also shows outstanding real-time tracking speed. On the Nvidia Tesla T4-16GB GPU, the FPS of CTIFTrack reaches 71.98. The code and demo materials will be available at https://github.com/vpsg-research/CTIFTrack.

Rubbing-induced site-selective deposition of 2D material patterns on nanomembranes

2D-layered materials (e.g., graphene and transition metal dichalcogenides) have attracted huge attention due to their unique mechanical and electrical properties. Emerging research efforts, which seek to combine device characterization and high-resolution electron micrography analysis for 2D-layered device features, demand nano/microlithographic techniques capable of producing ordered 2D material patterns on ultrathin membranes with nanoscale thicknesses. However, such membranes are so fragile that most conventional lithographic techniques can be hardly performed on them to generate 2D material patterns. Our previous works have demonstrated that the rubbing-induced site-selective (RISS) deposition method can produce arbitrary 2D semiconductor (e.g., MoS2 and Bi2Se3) patterns on regular device substrates. This fabrication route prevents the vulnerable 2D-layered structures from the detrimental damage introduced by plasma etching and resist-based lithography processes. In this work, we explore the applicability of RISS for directly producing 2D material patterns on nanomembranes. Specifically, this work shows that a polymeric interfacing layer on the rubbing template features, which can effectively prevent stress concentration during the rubbing process, is crucial to successful implementation of RISS processes on nanomembranes. Furthermore, we carried out the mechanics simulation of the Von Mises stress and pressure distribution on the RISS-processed membrane to identify the optimal rubbing load, which can generate sufficient triboelectric charge for material deposition but no damage to the membrane. Using this approach, we have successfully demonstrated the deposition of Bi2Se3 patterns on 25 nm SiOx nanomembranes and high-resolution transmission electron micrography characterization of the crystallographic structures.

Pyramid quaternion discrete cosine transform based ConvNet for cancelable face recognition

The current face scanning era can quickly and conveniently attain identity authentication, but face images imply sensitive information simultaneously. Under such context, we introduce a novel cancelable face recognition methodology by using quaternion transform based convolutional network. Firstly, face images in different modalities (e.g., RGB and depth or near-infrared) are encoded into full quaternion matrix for synchronous processing. Based on the designed multiresolution quaternion singular value decomposition, we can obtain pyramid representation. Then they are transformed through random projection for making the process noninvertible. Even if the feature template is compromised, a new one can be generated. Subsequently, a three-stream convolutional network is developed to learn features, where predefined filters are stemmed from quaternion two-dimensional discrete cosine transform basis. Extensive experiments on the TIII-D, NVIE and CASIA datasets have demonstrated that the proposed method obtains competitive performance, also satisfies redistributable and irreversible.

Transformer object tracking algorithm based on spatiotemporal template focal attention

Existing visual object tracking methods only use the target area of the first frame as a template, which makes it easy to fail in fast-changing and complex backgrounds. To solve this problem, this paper proposes a Transformer-based object tracking algorithm that focuses on the target focus information in the template and dynamically updates the template features. In order to reduce the interference of background information on attention, this algorithm uses a sparse Transformer module to achieve the interaction of feature information; a template focus attention module is also proposed to focus on the dynamic template features and the initial template features to retain the highly reliable feature information in the initial template. Experimental results show that the success rate and accuracy of this algorithm in the 0TB100 benchmark test are and respectively 70.9%, 91.6%and the success rate and accuracy are improved by 4.11%and respectively compared with the similar template update algorithm STARK 3.27%. This algorithm effectively addresses the limitations of existing visual object tracking methods and improves the accuracy and robustness of tracking.

UAT:Unsupervised object tracking based on graph attention information embedding

An excellent unsupervised tracker includes a powerful base tracker and an effective unsupervised tracking strategy. However, most base trackers lack internal feature representations for information embedding processes. Most unsupervised trackers are not robust enough in complex environments and lack an effective template update strategy. We propose an unsupervised object tracking based on graph attention information embedding (UAT) to solve the above problems. UAT combines graph attention mechanism with multi-scale features to construct a multi-scale graph attention module (MGA). MGA module dynamically and efficiently completes the information embedding between the template branch and the search area branch. The response map obtained by fusing the feature maps of the two branches is more informative about the location of the target. An attention based information reinforcement update module (RUM) improves the robustness of the tracker. RUM enhances the representation of the feature map in both the spatial dimension and the channel dimension. Template features are also updated indirectly through information transfer between the two branches. RUM suppresses background interference and improves network perception during tracking. Experiments on challenging benchmarks such as VOT2018, VOT2019, TrackingNet, OTB100, LaSOT and UAV123 demonstrate that the proposed UAT achieves state-of-the-art performance in unsupervised trackers.

Attentional templates for target features versus locations

Visual search is guided by visual working memory representations (i.e., attentional templates) that are activated prior to search and contain target-defining features (e.g., color). In the present study, we tested whether attentional templates can also contain spatial target properties (knowing where to look for) and whether attentional selection guided by such feature-specific templates is equally efficient than selection that is based on feature-specific templates (knowing what to look for). In every trial, search displays were either preceded by semantic color or location cues, indicating the upcoming target color or location, respectively. Qualitative differences between feature- and location-based template guidance were substantiated in terms of selection efficiency in low-load (one target color/location) versus high-load trials (two target colors/locations). Behavioral and electrophysiological (N2pc) measures of target selection speed and accuracy were combined for converging evidence. In line with previous studies, we found that color search was highly efficient, even under high-low conditions, when multiple attentional templates were activated to guide attentional selection in a spatially global fashion. Importantly, results in the location task almost perfectly mirrored the findings of the color task, suggesting that multiple templates for different target locations were activated concurrently when two possible target locations were task relevant. Our findings align with accounts that assume a common neuronal network during preparation for location and color search, but regard spatial and feature-based selection mechanisms as independent.

Three-Dimensional Printing Technology Based on Digital Orthopedics: 5-Point Positioning Point-Contact Pedicle Navigation Template in the Case of Scoliosis and Complex Pedicle.

Imperfect fitting of the navigation template leads to prolonged surgery time and increased blood loss. These problems have not been effectively addressed in previous research. This study explores the efficacy of a novel 5-point positioning point-contact pedicle navigation template in complex pedicle situations in scoliosis. This study employed a retrospective controlled design. From November 2019 to November 2023, 28 patients with scoliosis and complex pedicle were selected and underwent scoliosis correction surgery. A 5-point positioning point-contact pedicle navigation template was used intraoperatively to guide pedicle screw placement. Matched with 56 historical cases as a control group. The analysis included screw placement time, screw placement bleeding volume, fluoroscopy frequency, manual repositioning frequency, screw placement accuracy and grade, screw placement complications, and main curve correction rate. Continuous variables were compared using the independent samples t-test. Categorical data were analyzed with the chi-square test. All 28 patients successfully underwent surgery, with a total of 268 pedicle screws placed. The surgery duration ranged from 220 to 410 min, with an average of (283.16 ± 51.26) min. Intraoperative blood loss ranged from 630 to 1900 mL, with an average of (902.17 ± 361.25) mL. Pedicle screw placement time ranged from 60 to 130 min, with an average of (85.24 ± 24.65) min. Pedicle screw placement bleeding volume ranged from 40 to 180 mL, with an average of (76.47 ± 42.65) mL. Fluoroscopy frequency ranged from 3 to 7 times, with an average of (4.31 ± 1.14) times. Manual repositioning frequency ranged from 0 to 2 times, with an average of (0.46 ± 0.58) times. Pedicle screw placement grades: Grade I: 237 screws; Grade II: 25 screws; Grade III: 6 screws; Grade IV: 0 screws. There were no screw-related complications. The correction rate ranged from 46% to 68%, with an average of (55.83 ± 9.22)%. Compared to the experienced screw group, the differences in screw placement time, screw placement bleeding volume, fluoroscopy procedures, and manual redirections were statistically significant (p < 0.05). The 5-point positioning point-contact pedicle navigation template features a claw-like structure that securely adapts to various deformed vertebral facet joints, avoiding drift phenomena and ensuring accurate screw placement. Its pointed contact structure with the lamina of the spine avoids extensive and complete detachment of posterior structures, reducing blood loss, surgery time, and trauma. Predesigned pedicle screw entry points and directions reduce fluoroscopy frequency and surgery time.

Siamese Tracking Network with Multi-attention Mechanism

Object trackers based on Siamese networks view tracking as a similarity-matching process. However, the correlation operation operates as a local linear matching process, limiting the tracker’s ability to capture the intricate nonlinear relationship between the template and search region branches. Moreover, most trackers don’t update the template and often use the first frame of an image as the initial template, which will easily lead to poor tracking performance of the algorithm when facing instances of deformation, scale variation, and occlusion of the tracking target. To this end, we propose a Simases tracking network with a multi-attention mechanism, including a template branch and a search branch. To adapt to changes in target appearance, we integrate dynamic templates and multi-attention mechanisms in the template branch to obtain more effective feature representation by fusing the features of initial templates and dynamic templates. To enhance the robustness of the tracking model, we utilize a multi-attention mechanism in the search branch that shares weights with the template branch to obtain multi-scale feature representation by fusing search region features at different scales. In addition, we design a lightweight and simple feature fusion mechanism, in which the Transformer encoder structure is utilized to fuse the information of the template area and search area, and the dynamic template is updated online based on confidence. Experimental results on publicly tracking datasets show that the proposed method achieves competitive results compared to several state-of-the-art trackers.

Recurrent inhibition refines mental templates to optimize perceptual decisions.

Translating sensory inputs to perceptual decisions relies on building internal representations of features critical for solving complex tasks. Yet, we still lack a mechanistic account of how the brain forms these mental templates of task-relevant features to optimize decision-making. Here, we provide evidence for recurrent inhibition: an experience-dependent plasticity mechanism that refines mental templates by enhancing γ-aminobutyric acid (GABA)-mediated (GABAergic) inhibition and recurrent processing in superficial visual cortex layers. We combine ultrahigh-field (7 T) functional magnetic resonance imaging at submillimeter resolution with magnetic resonance spectroscopy to investigate the fine-scale functional and neurochemical plasticity mechanisms for optimized perceptual decisions. We demonstrate that GABAergic inhibition increases following training on a visual (i.e., fine orientation) discrimination task, enhancing the discriminability of orientation representations in superficial visual cortex layers that are known to support recurrent processing. Modeling functional and neurochemical plasticity interactions reveals that recurrent inhibitory processing optimizes brain computations for perpetual decisions and adaptive behavior.

Identification of microalgal particles using pixel feature analysis of Mueller matrix images

Fine sorting of aquatic particles is of great significance for water environment monitoring. Natural water contains aquatic particles that exhibit a high degree of diversity and complexity, and the identification of aquatic particles remains a persistent challenge in the field. In this article, we propose a new technique for identifying the target species of microalgal particles by using the pixel feature analysis of Mueller matrix images. This technique is independent of any prior knowledge or data about the existing particles in the environment, which is advantageous when applied to real-world situations. The pixel-level polarimetric features are fully leveraged to construct polarization feature templates, which can be used to characterize and filter specific microalgal particles in complex environments. This method could enable the accurate detection of harmful algal blooms species in natural water, which can facilitate early warning of algal blooms. The preliminary results show that the recall rate reached 97.2%, and the average accuracy is 98.9%, which demonstrate the effectiveness of this approach for identifying the target species of aquatic particles in natural water.

An efficient object tracking based on multi‐head cross‐attention transformer

AbstractObject tracking is an essential component of computer vision and plays a significant role in various practical applications. Recently, transformer‐based trackers have become the predominant method for tracking due to their robustness and efficiency. However, existing transformer‐based trackers typically focus solely on the template features, neglecting the interactions between the search features and the template features during the tracking process. To address this issue, this article introduces a multi‐head cross‐attention transformer for visual tracking (MCTT), which effectively enhance the interaction between the template branch and the search branch, enabling the tracker to prioritize discriminative feature. Additionally, an auxiliary segmentation mask head has been designed to produce a pixel‐level feature representation, enhancing and tracking accuracy by predicting a set of binary masks. Comprehensive experiments have been performed on benchmark datasets, such as LaSOT, GOT‐10k, UAV123 and TrackingNet using various advanced methods, demonstrating that our approach achieves promising tracking performance. MCTT achieves an AO score of 72.8 on the GOT‐10k.

Constructing the KOR152 Korean Young Adult Brain Atlas Utilizing the State-of-the-Art Method for the Age-Specific Population.

Spatial normalization is an essential process for comparative analyses that heavily depends on the standard brain template used. Brain morphological differences are observed in different populations due to genetic and environmental factors, causing mismatches in regions when the data are normalized to different population templates. Recent studies have indicated differences between Caucasian and East Asian populations as well as within East Asian populations, suggesting the necessity of population-specific brain templates. Thus, this study aimed to construct a Korean young adult age-specific brain template utilizing an advanced method of template construction to update the currently available Korean template. The KOR152 template was constructed via affine and nonlinear iterative procedures based on prior studies. We compared the morphological features of different population templates (MNI152, Indian_157, and CN200). The distance and volumetric changes before and after registering the data to these templates were calculated for registration accuracy. The KOR152 global brain features revealed a shorter overall length than the other population templates. The registration accuracy by distance and volumetric change was significantly lower than that of the other population templates, implying that the KOR152 was more accurate than other templates for the young adult Korean population. This study provided evidence for the need for a population-specific template that may be more appropriate for structural and functional studies in Korean populations.

CDTracker: Coarse-to-Fine Feature Matching and Point Densification for 3D Single-Object Tracking

Three-dimensional (3D) single-object tracking (3D SOT) is a fundamental yet not well-solved problem in 3D vision, where the complexity of feature matching and the sparsity of point clouds pose significant challenges. To handle abrupt changes in appearance features and sparse point clouds, we propose a novel 3D SOT network, dubbed CDTracker. It leverages both cosine similarity and an attention mechanism to enhance the robustness of feature matching. By combining similarity embedding and attention assignment, CDTracker performs template and search area feature matching in a coarse-to-fine manner. Additionally, CDTracker addresses the problem of sparse point clouds, which commonly leads to inaccurate tracking. It incorporates relatively dense sampling based on the concept of point cloud segmentation to retain more target points, leading to improved localization accuracy. Extensive experiments on both the KITTI and Waymo datasets demonstrate clear improvements in CDTracker over its competitors.

Bayesian inference methodology for primordial power spectrum reconstructions from Large Scale Structure

We use Bayesian inference to develop a non-parametric method to reconstruct the primordial power spectrum Pℛ (k) from Large Scale Structure (LSS) data. The performance of the method is assessed by testing it against simulations of the clustering of high-z (QSOs) objects. Their clustering is derived from different templates of the primordial power spectrum motivated by models of inflation: the Standard Model power law characterized by the two parameters As and ns ; a local feature template; and a global oscillatory template. The primordial power spectrum is reconstructed using N knots in the log {k, Pℛ (k)} plane while sampling the cosmological parameters {H 0, Ω b , Ω c }. We use two statistical tests to examine the reconstructions for signs of primordial features: a global test comparing the evidences and a novel local test quantifying the power of the hypothesis test between the power law model and the marginalized probability over N model. We also discuss results of an application to low-z (ELGs) objects with two different photometric errors keeping the cosmology fixed. The method shows good performance in all scenarios considered. In particular, the tests show no feature detection for the standard power-law primordial power spectrum; yet, the method is able to detect power spectrum deviations at a percent level for all considered features, combining either the low-z or the high-z redshift bins. In addition, we include a test proof-of-concept application to real data from the Sloan Digital Sky Survey Luminous Red Galaxy Data Release 4 (SDSS LRG 04), finding no preference for deviations from the primordial power law. The method is flexible, model independent, and suitable for its application to existing and future LSS surveys.

ДОСЛІДЖЕННЯ ІНТЕЛЕКТУАЛЬНИХ МОДЕЛЕЙ КЛАСИФІКАЦІЇ НЕВИЗНАЧЕНИХ ДАНИХ З ВИМОГАМИ ДОСТОВІРНОСТІ РЕЗУЛЬТАТІВ

The report contains the results of research and development of intellectual models for managing complex systems under conditions of data uncertainty based on classification procedures, which ensure reliable solution of tasks considering the assessment of the boundary dimensionality of models. The possibilities of improving Hamming neural networks for data classification in formats of fuzzy variables and certainty factor CF(A) are investigated. The features of the mathematical model of classification tasks based on a set of feature templates are determined. A software complex of information technology for assignment/selection of performers, as well as determination of authorship of Ukrainian-language works based on classification of sets of templates with certain fuzzy features, is presented. The software complex utilizes the reduction and Cohen's kappa statistics procedures proposed in the report.

Deep secure PalmNet: A novel cancelable palmprint template protection scheme with deep attention net and randomized hashing security mechanism

Overcoming challenges in the current cancelable palmprint template is essential for enhancing privacy and security in biometric systems, as these templates exhibit vulnerabilities in both template security and recognition performance. To address these challenges, we present a novel cancelable palmprint template protection scheme with a deep attention net and randomized hashing security mechanism. Firstly, to enhance recognition performance, we designed a deep attention net that is integrated into the model, aiming to improve its feature extraction ability. Additionally, in recognition of the paramount importance of security and privacy, we introduce a randomized hashing security mechanism. This mechanism, incorporating chaotic sequences as weights in a neuron activation layer, appends to the model and enables dynamic control of neuron activation while generating diverse palmprint feature templates. Subsequently, it enhances security and privacy by combining a Logistic-Tent-Sine (LTTS) random key with palmprint feature values through matrix multiplication. Furthermore, to optimize efficiency, particularly with a large dataset, a binarization layer is implemented using the Straight-Through Estimation (STE) algorithm. This layer contributes to computational efficiency and expedient data processing, further improving the performance of our palmprint template protection. The experimental results validate the outstanding accuracy of this scheme on TJU and PolyU palmprint datasets, establishing it as a state-of-the-art solution with remarkable recognition performance. Moreover, security analysis confirms its compliance with cancelable biometric template protection criteria, ensuring superior irreversibility, unlinkability, revocability, and privacy against various attacks.