Flexible Adaptation Research Articles

Hyperspectral image super resolution using deep internal and self‐supervised learning

AbstractBy automatically learning the priors embedded in images with powerful modelling capabilities, deep learning‐based algorithms have recently made considerable progress in reconstructing the high‐resolution hyperspectral (HR‐HS) image. With previously collected large‐amount of external data, these methods are intuitively realised under the full supervision of the ground‐truth data. Thus, the database construction in merging the low‐resolution (LR) HS (LR‐HS) and HR multispectral (MS) or RGB image research paradigm, commonly named as HSI SR, requires collecting corresponding training triplets: HR‐MS (RGB), LR‐HS and HR‐HS image simultaneously, and often faces difficulties in reality. The learned models with the training datasets collected simultaneously under controlled conditions may significantly degrade the HSI super‐resolved performance to the real images captured under diverse environments. To handle the above‐mentioned limitations, the authors propose to leverage the deep internal and self‐supervised learning to solve the HSI SR problem. The authors advocate that it is possible to train a specific CNN model at test time, called as deep internal learning (DIL), by on‐line preparing the training triplet samples from the observed LR‐HS/HR‐MS (or RGB) images and the down‐sampled LR‐HS version. However, the number of the training triplets extracted solely from the transformed data of the observation itself is extremely few particularly for the HSI SR tasks with large spatial upscale factors, which would result in limited reconstruction performance. To solve this problem, the authors further exploit deep self‐supervised learning (DSL) by considering the observations as the unlabelled training samples. Specifically, the degradation modules inside the network were elaborated to realise the spatial and spectral down‐sampling procedures for transforming the generated HR‐HS estimation to the high‐resolution RGB/LR‐HS approximation, and then the reconstruction errors of the observations were formulated for measuring the network modelling performance. By consolidating the DIL and DSL into a unified deep framework, the authors construct a more robust HSI SR method without any prior training and have great potential of flexible adaptation to different settings per observation. To verify the effectiveness of the proposed approach, extensive experiments have been conducted on two benchmark HS datasets, including the CAVE and Harvard datasets, and demonstrate the great performance gain of the proposed method over the state‐of‐the‐art methods.

A latent profile analysis of resilience and the associations with flourishing in emergency nurses exposed to workplace violence in China.

Workplace violence (WPV) exposure has been particularly prevalent in emergency nurses, coupled with negative mental health outcomes. Few explored resilience after WPV exposure using the perspective of positive mental health. We aimed to identify latent profiles of resilience and examine associations with flourishing in emergency nurses with WPV exposure. A total of 1241 Chinese emergency nurses were surveyed on the frequency of WPV exposure, resilience, and flourishing. Eight hundred and twenty five participants (90.1% female, 86.6% aged 20-39) reported WPV exposure at least once in the past 3 months. Latent profile analysis identified profiles of resilience as low resilience (15.0%), moderate decision respond and interpersonal link with low rational thought and flexible adaption (18.0%), moderate resilience (31.0%), high decision respond, interpersonal link, and rational thought with moderate flexible adaption (16.0%), and high resilience (20.0%). Higher flourishing was observed in profiles of moderate resilience, high decision respond, interpersonal link, and rational thought with moderate flexible adaption, and high resilience. We differentiated profiles of resilience and the associations with flourishing in emergency nurses with WPV exposure, which informed targeted interventions for promoting positive mental health.

STRATEGIC DECISION MAKING IN A HETEROGENEOUS AND UNCERTAIN BUSINESS ENVIRONMENT: INTEGRATED MODELS FOR ANALYSIS AND FORECASTING

This article provides a comprehensive examination of models employed in strategic decision-making within a context characterized by heterogeneity and uncertainty in the business environment. It elucidates prominent theories and methodologies utilized for the identification, analysis, and response to diverse economic, social, and technological factors influencing the market environment. The discourse encompasses the evolution in strategic decision-making paradigms, encompassing transitions towards scenario planning, the delicate equilibrium between intuitive approaches and analytical methodologies, and the organizational adaptation to internal variations. The research methodology proposed herein is grounded in a systematic approach, emphasizing analysis and synthesis. The article directs its focus towards diverse integrative analysis models that amalgamate internal and external facets of organizations, facilitating efficacious strategic decision-making. The scrutiny of models such as the Ansoff Matrix, Porter's Five Forces, McKinsey 7S Model, and BCG Growth-Share Matrix reveals that each model possesses distinct advantages and limitations contingent upon the specific conditions and requisites of the business environment in question. The author asserts that there exists no singular optimal model universally applicable to all scenarios, underscoring the imperative for flexible integration and adaptation of varied approaches. Illustratively, this may involve the amalgamation of Porter's Five Forces model with SWOT analysis, accounting for both internal and external factors, or the augmentation of the McKinsey 7S Model with external analyses, such as PESTEL analysis. The conclusions drawn in the article underscore that the achievement of success in strategic decision-making within the contemporary dynamic and unpredictable business environment hinges on organizations' capacity for flexibility, adaptability, and innovation. The escalating complexity and rapid pace of change necessitate companies to formulate comprehensive strategies that encompass various aspects and approaches for the development of effective solutions.

Beyond the Classroom in Indonesia: How Organizational Support Shapes Academic Stress and Emotional Wellness

In the era of modern education characterized by technological developments, understanding the factors that affect the academic load and emotional well-being of lecturers and students is very important. This study explores the impact of organizational support, time flexibility, and technological adaptation on the academic load and emotional well-being of lecturers and students in the college environment. Data was collected through an online questionnaire from 120 respondents, consisting of 92 female and 28 male respondents. Meanwhile, from the point of view of time flexibility, as many as 60 respondents came from universities that apply time flexibility. In comparison, the other 60 respondents came from universities that do not apply time flexibility. The data analysis technique used in this study is Partial Least Squares Structural Equation Modeling (PLS-SEM) with SmartPLS software. The analysis results show that high awareness and understanding of lecturers and students on academic load is the main factor that eliminates the significant influence of variables such as organizational support, time flexibility, and technological adaptation to academic load. Nonetheless, the focus of the study shows that organizational support, technological adaptation, and academic load significantly influence emotional well-being. A deep understanding of the relationship between these variables can assist universities in creating an academic environment conducive to the development and well-being of the academic community.

ВПЛИВ НИЗЬКОЧАСТОТНОГО ЕЛЕКТРОМАГНІТНОГО ВИПРОМІНЮВАННЯ НА СИСТЕМУ КОМПЛЕМЕНТУ (ЕКСПЕРИМЕНТАЛЬНЕ ДОСЛІДЖЕННЯ)

Abstract. The article presents the data of an experimental study on determining the state of the complement system under the conditions of exposure to low-frequency electromagnetic radiation (EMR) on the body of rats under the conditions of a subacute experiment. The relevance of the research is dictated by the fact that low-frequency EMR can change the membrane potential and disrupt nerve cell conduction. This can primarily affect the functioning of the hypothalamic-pituitary-adrenal system, which also influences the effectiveness of the immune system functioning. In the conditions of the natural experiment, male rats were exposed to radiation of 600 V/m with a frequency of 70 kHz for 4 hours every day for 30 days. The studies of immunological indices were carried out in the dynamics of the experiment on the 5th, 15th and 30th days. The study showed that under conditions of 30-day exposure to low-frequency EMR, the humoral link of immunity was characterized by suppression of the C4 component in the first half of the study, followed by a compensatory increase in the C4, C5 components. It has been proven that the general effector reactions of activation of the complement system are able to influence the functions of lymphocytes and macrophages. Thus, under the conditions of an anthropogenic factor, a flexible adaptation process takes place in the body of rats, which involves the redistribution of protein functions for the purpose of "economic" functioning of the body under the conditions of stressor action.

An Untethered Soft Crawling Robot Driven by Wireless Power Transfer Technology.

Soft robots based on flexible materials have attracted the attention due to high flexibility and great environmental adaptability. Among the common driving modes, electricity, light, and magnetism have the limitations of wiring, poor penetration capability, and sophisticated equipment, respectively. Here, an emerging wireless driving mode is proposed for the soft crawling robot based on wireless power transfer (WPT) technology. The receiving coil at the robot's tail, as an energy transfer station, receives energy from the transmitting coil and supplies the electrothermal responsiveness to drive the robot's crawling. By regulating the WPT's duration to control the friction between the robot and the ground, bidirectional crawling is realized. Furthermore, the receiving coil is also employed as a sensory organ to equip the robot with localization, ID recognition, and sensing capabilities based on electromagnetic coupling. This work provides an innovative and promising strategy for the design and integration of soft crawling robots, exhibiting great potential in the field of intelligent robots.

Methodological recommendations on the practical application of tools for ensuring safe information interaction of cluster members

Based on the conducted research, the article presents methodological recommendations on the choice of implementation of the coordination organizational and management body of the cluster in the aspect of its information and communication interaction, depending on the type of cluster. It also describes proposals for flexible adaptation of cluster employees based on change management, aimed at overcoming resistance to innovations. The recommendations described in the article will make it possible to effectively implement in Russian clusters tools for automating the processing of requests to determine the authenticity of the subject, extraction and parallel processing of the targeted information pool while maintaining its speed, while optimally revealing their advantages in each practical case, realizing synergetic and multiplicative effects, as well as obtaining opportunities to consolidate the result and subsequent replication of such experience.

Fishing methods of Can Gio fishermen: tradition and transformation

Can Gio is the only coastal district in Ho Chi Minh City, where its natural conditions are favorable for fishing by a wide range of methods and gears. Therefore, from its ancient days, in addition to agricultural activities, fishing became a means of livelihood for many households in Can Gio. In this article, we have deployed the field research method to study common fishing gears frequently used by fishermen in Can Gio district in the past and present to catch fish in “shallow water and deep water”, specifically, stow nets, other typical nets, hooks and lines, types of rakes, etc. Some fishing gears are still used by Can Gio fishermen today, but have changed adapting to the current conditions, especially in terms of crafting materials as well as fishing areas. The research on the change in fishing gears in Can Gio not only focuses on how fish are caught but also reflects the local fishermen community’s flexible adaptation through their fishing over time. More importantly, this greatly contributes to maintaining and fostering the traditional value of fishing in Can Gio district not only at the present time but also in the future.

Visit planner: A personalized mobile trip design application based on a hybrid recommendation model

The paper presents Visit Planner (ViP), a mobile application prototype that provides a solution to the challenging tourist trip design problem. ViP follows a holistic approach offering personalized recommendations for Points of Interest (POIs) based on preferences either explicitly collected by the application, or inferred by the users? ongoing interaction with the system. ViP proposes to the final user, a trajectory of POIs calculated using an Expectation Maximization method that maximizes user satisfaction taking into consideration a variety of time and spatial constraints for both users and POIs. Additionally, POIs are divided into categories, so that a certain number of POIs from each category to be included in the final itinerary. The application is implemented as a user-interactive system that allows the flexibility for easy content adaptation and facilitates management of content and services by the user. The prototype has been implemented for Android-based smartphones, on an open application environment, using standard communication protocols and open database technology. Currently, it is applied to the city of Agios Nikolaos in Crete, and is available for download from Google play.

Acceptability and feasibility of a brief intervention to enhance resilience among young people and their families in India and Kenya.

Enhancing resilience is one way to prevent future mental illnesses and encourage recovery in the face of adversity. To develop and test the acceptability and feasibility (A&F) of a combined family and individual resilience intervention in two rural/semi-rural low-income settings in India and Kenya. We developed a five-session intervention including Life Skills Education (LSE) and a model of family resiliency. Among adolescents aged 14-16 years and their families in India and Kenya, we collected socio-demographics and audio records of delivery and undertook a process evaluation. Due to COVID-19, we developed a hybrid intervention. The facilitators and participants preferred the in-person model. India: Of 17 families, 10 fully completed the intervention. They identified three critical components: 1) story-telling, 2) cooperation and working together and 3) expressing feelings. Kenya: All 15 families completed the intervention. Critical elements were 1) seeing social value in learning to make good decisions, 2) promoting an optimistic view of life, 3) hearing stories that resonated with their situation and 4) enhancing family performance through knowledge-building. We mapped the active ingredients, showing fidelity and acceptability. The intervention showed promising A&F parameters. Flexibility and local adaptation were important for delivery.

Core–shell Pd(0)@His-SiO2/CoFe2O4 nano-composite as a magnetically recoverable heterogeneous catalyst for the deprotection of oximes and Heck coupling

Rational engineering of core–shell nanostructure-based catalysts have received significant attention owing to their potential for exhibiting unique properties such as durability, structural flexibility, and porous shell adaptability.

A Nonlinear PID-Incorporated Adaptive Stochastic Gradient Descent Algorithm for Latent Factor Analysis

High-dimensional and incomplete (HDI) data are commonly encountered in various big data-related applications concerning the complex interactions among numerous nodes, such as the user-item iterations in a recommender system. A stochastic gradient descent (SGD)-based latent factor analysis (LFA) model can perform efficient representation learning to such HDI data, thereby extracting useful knowledge from them. However, a standard SGD algorithm updates a latent factor based on the current stochastic gradient only, without the considerations on the past information, making a resultant model suffer from slow convergence. To address this critical issue, this paper proposes an Adaptive Non-linear PID-incorporated SGD (ANPS) algorithm with two-fold ideas: 1) rebuilding the instant learning error when computing the stochastic gradient following the principle of a nonlinear PID controller to incorporate past update information into the learning scheme efficiently, and 2) implementing gain parameter adaptation following the principle of particle swarm optimization (PSO). Experiments on six widely-adopted HDI datasets demonstrate that compared with state-of-the-art LFA models, an ANPS-based LFA model achieves significant advantage in both efficiency and accuracy. Moreover, its flexible gain parameter adaptation mechanism greatly boosts its practicability for real issues. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —In many industrial applications like recommender systems, social network systems, and cloud service systems, people usually encounter numerous nodes and their highly-incomplete relationships. An HDI matrix is commonly adopted to describe such specific relationships. One of the major challenges is to acquire useful knowledge from an HDI matrix efficiently and accurately for various data analysis tasks, e.g., accurate recommendation, community detection, and web service selection. An SGD-based LFA model has been widely adopted to tackle this issue. However, it suffers from slow convergence that leads to considerable time cost on large-scale datasets. This study proposes an ANPS algorithm following the principle of a nonlinear PID controller. With it, an ANPS-based LFA model is achieved, which possesses fast convergence rate on an industrial HDI matrix. The proposed ANPS algorithm can be leveraged for different types of various machine learning models, thereby improving their utility and scalability in practice.

Concepts and features of business scaling in the conditions of modern challenges

The article considered scientific approaches to the interpretation of the concept of «business scaling», reflecting its various aspects. It is noted that business scaling is a strategy because it requires certain coordinated actions and is implemented when the company is ready for scaling. It is determined that it can be aimed at achieving the following goals: increasing profits, capturing the market, gaining access to investments and expanding financing, developing technologies, ensuring competitiveness, improving the image, increasing the rating and influence of the company, increasing brand awareness, etc. It is noted that the main measures to achieve these scaling goals are quantitative and (or) qualitative changes in the functioning of the company. The author's definition of this term is formulated as a flexible, agile, quickly adaptive strategy with constant improvement of business processes to achieve strategic goals. The main features of such a strategy in the conditions of modern challenges have been revealed: the increase in the scale of business does not affect costs in the same way; the increase in income occurs faster than the increase in expenses, which ensures an increase in profit, but in the first stages of the implementation of this strategy, you can face an increase in expenses without an increase in income; scaling creates additional business risks, which leads to increased costs and requires increased control over business processes, etc. The components of this strategy are determined: a high-quality and mature product, the demand for which significantly exceeds the supply, sufficient margin for all participants, reliable partners, and a clear business model based on the optimization of internal and external business processes. It is noted that the basis of effective scaling is: the introduction of innovations; improvement of the marketing component; systematic and qualitative analysis of markets, company potential and expansion opportunities. It was established that such a strategy requires: the creation of an effective and optimal organizational structure; communication systems; changes in management approaches; motivation and synchronization of the work of the project team; involvement of business development specialists; delegation of powers and management functions. It was found that business scaling in the face of modern challenges is not only a source of its growth, but thanks to the flexibility, maneuverability, and quick adaptation of this strategy, it is also a guarantee of protecting the company from adverse changes. Even if the project (business) is local, a successfully found niche will make it possible to make it large-scale.

Етапи та інструменти управління стратегічними змінами на підприємствах під час кризових ситуацій

In the current conditions of doing business in Ukraine, the importance of adapting the system of strategic change management to the complex socioeconomic conditions faced by Ukrainian enterprises is sharply increasing. Accordingly, the way in which enterprises can effectively respond to exogenous (external) challenges, in particular to crisis situations that require prompt and flexible strategic decisions to ensure their survival and further development, is of particular relevance. The aim of this article is to substantiate the theoretical provisions of management of strategic changes in the enterprise and, on this basis, to develop effective instruments for managing them in crisis situations at enterprises. Various approaches to change management are analyzed, which can be used to plan and implement the necessary changes in the strategy and operational activities of domestic enterprises in the context of combating and overcoming the crisis. Particular attention is paid to the aspects of leadership and corporate culture as key factors in successful adaptation to change. The importance of involving employees in the change management process, the need to create effective communication within the organization and the introduction of a flexible, adaptive corporate culture that promotes innovation are emphasized. In the context of the Ukrainian economy, attention is focused on the specifics of management during the crisis, taking into account the military, political, social and economic instability in the country. An instrumentarium of strategic change management, adapted to Ukrainian realities, including analysis of the existing crisis business environment, strategic planning, prompt implementation of changes and monitoring of results, has been proposed. It is emphasized that flexibility, forecasting and rapid adaptation to changes are key to ensuring the stability and development of enterprises in Ukraine in conditions of unpredictability and crises. The proposed research can be useful for researchers in the field of strategic change management, analysts, business associations, enterprises in Ukraine and abroad. Further research in this direction will be aimed at the development of methods for managing strategic changes in enterprises, which are based on the observation of rapid changes in the exogenous and endogenous environment of the enterprise.

Research on control strategy of pneumatic soft bionic robot based on improved CPG.

To achieve the accuracy and anti-interference of the motion control of the soft robot more effectively, the motion control strategy of the pneumatic soft bionic robot based on the improved Central Pattern Generator (CPG) is proposed. According to the structure and motion characteristics of the robot, a two-layer neural network topology model for the robot is constructed by coupling 22 Hopfield neuron nonlinear oscillators. Then, based on the Adaptive Neuro-Fuzzy Inference System (ANFIS), the membership functions are offline learned and trained to construct the CPG-ANFIS-PID motion control strategy for the robot. Through simulation research on the impact of CPG-ANFIS-PID input parameters on the swimming performance of the robot, it is verified that the control strategy can quickly respond to input parameter changes between different swimming modes, and stably output smooth and continuous dynamic position signals, which has certain advantages. Then, the motion performance of the robot prototype is analyzed experimentally and compared with the simulation results. The results show that the CPG-ANFIS-PID motion control strategy can output coupled waveform signals stably, and control the executing mechanisms of the pneumatic soft bionic robot to achieve biological rhythms motion propulsion waveforms, confirming that the control strategy has accuracy and anti-interference characteristics, and enable the robot have certain maneuverability, flexibility, and environmental adaptability. The significance of this work lies in establishing a CPG-ANFIS-PID control strategy applicable to pneumatic soft bionic robot and proposing a rhythmic motion control method applicable to pneumatic soft bionic robot.

Impaired flexible reward learning in ADHD patients is associated with blunted reinforcement sensitivity and neural signals in ventral striatum and parietal cortex

Reward-based learning and decision-making are prime candidates to understand symptoms of attention deficit hyperactivity disorder (ADHD). However, only limited evidence is available regarding the neurocomputational underpinnings of the alterations seen in ADHD. This concerns flexible behavioral adaption in dynamically changing environments, which is challenging for individuals with ADHD. One previous study points to elevated choice switching in adolescent ADHD, which was accompanied by disrupted learning signals in medial prefrontal cortex.Here, we investigated young adults with ADHD (n = 17) as compared to age- and sex-matched controls (n = 17) using a probabilistic reversal learning experiment during functional magnetic resonance imaging (fMRI). The task requires continuous learning to guide flexible behavioral adaptation to changing reward contingencies. To disentangle the neurocomputational underpinnings of the behavioral data, we used reinforcement learning (RL) models, which informed the analysis of fMRI data.ADHD patients performed worse than controls particularly in trials before reversals, i.e., when reward contingencies were stable. This pattern resulted from ‘noisy’ choice switching regardless of previous feedback. RL modelling showed decreased reinforcement sensitivity and enhanced learning rates for negative feedback in ADHD patients. At the neural level, this was reflected in a diminished representation of choice probability in the left posterior parietal cortex in ADHD. Moreover, modelling showed a marginal reduction of learning about the unchosen option, which was paralleled by a marginal reduction in learning signals incorporating the unchosen option in the left ventral striatum.Taken together, we show that impaired flexible behavior in ADHD is due to excessive choice switching (‘hyper-flexibility’), which can be detrimental or beneficial depending on the learning environment. Computationally, this resulted from blunted sensitivity to reinforcement of which we detected neural correlates in the attention-control network, specifically in the parietal cortex. These neurocomputational findings remain preliminary due to the relatively small sample size.

Primary care-based screening and recruitment for an adolescent depression prevention trial: Contextual considerations during a youth mental health crisis.

Rising rates of adolescent depression in the wake of COVID-19 and a youth mental health crisis highlight the urgent need for accessible mental healthcare and prevention within primary care. Digital mental health interventions (DMHIs) may increase access for underserved populations. However, these interventions are not well studied in adolescents, nor healthcare settings. The purpose of this study was to identify barriers and facilitators to screening and recruitment activities for PATH 2 Purpose (P2P): Primary Care and Community-Based Prevention of Mental Disorders in Adolescents, a multi-site adolescent depression prevention trial comparing two digital prevention programs within four diverse health systems in two U.S. states. This qualitative study is a component of a larger Hybrid Type I trial. We conducted semi-structured key informant interviews with clinical and non-clinical implementers involved with screening and recruitment for the P2P trial. Informed by the Consolidated Framework for Implementation Research (CFIR), interviews were conducted at the midpoint of the trial to identify barriers, facilitators, and needed adaptations, and to gather information on determinants that may affect future implementation. Respondents perceived the P2P trial as valuable, well aligned with the mission of their health systems. However, several barriers were identified, many of which stemmed from influences outside of the healthcare settings. Universal and site-specific outer setting influences (COVID-19 pandemic, youth mental health crisis, local community conditions) interacted with Inner Setting and Innovation domains to create numerous challenges to the implementation of screening and recruitment. Our findings emphasize the need for ongoing, comprehensive assessment of dynamic inner and outer setting contexts prior to and during implementation of clinical trials, as well as flexibility for adaptation to unique clinical contexts. The CFIR is useful for assessing determinants during times of rapid inner and outer setting change, such as those brought on by the COVID-19 pandemic, youth mental health crisis, and the corresponding exacerbation of resource strain within healthcare settings. PATH 2 Purpose: Primary Care and Community-Based Prevention of Mental Disorders in Adolescents https://www.clinicaltrials.gov/study/NCT04290754.

Association Between Ego-Resilience and Social Isolation Among Urban Older People.

Social isolation among older adults affects their physical and mental health. Ego-resilience is associated with flexible adaptation to various environments and acceptable behaviors. However, its association with social isolation among older people is unclear. Therefore, a health survey was conducted with 510 adults aged 65 years or older to assess social isolation, its associated factors, and ego resilience. The results showed that the social isolation group had a lower median ego-resiliency scale score (42.0) compared to the non-social isolation group (38.0). The social isolation group were mostly males without spouses, took time to move, had more depressive moods, and poorer subjective health status. The non-social isolated group had faster maximum walking speed. Low ego-resilience was newly identified as a factor associated with social isolation among older people, as were being male, having no spouse, and depressive symptoms. Mediation analysis revealed that ego-resilience affected social isolation and was partially mediated by depressive symptoms.

THE IMPLEMENTATION OF MARKETING MANAGEMENT SOLUTIONS IN AGRICULTURAL ENTERPRISES

The article examines the components of the marketing activity management system and makes proposals for management decisions regarding the implementation of marketing tools in the activities of agricultural enterprises. It was determined that for agrarian enterprises, the use of marketing tools based on marketing management contributes to its flexible adaptation to changes in the environment, solving strategic tasks of further effective development and maintaining the level of competitiveness in the market. The authors present their own vision of the essence of marketing management and measures that are the basis of marketing activity management. It was established that marketing management at the enterprise is implemented using a set of marketing tools. Proposals regarding the adoption of effective management decisions for each component of the marketing complex are provided. Practical recommendations on the organization of the system of marketing activities for agricultural enterprises have been formed.

Refocus the Attention for Parameter-Efficient Thermal Infrared Object Tracking.

Introducing deep trackers to thermal infrared (TIR) tracking is hampered by the scarcity of large training datasets. To alleviate the predicament, a common approach is full fine-tuning (FFT) based on pretrained RGB parameters. Nevertheless, due to its inefficient training pattern and representation collapse risk, some parameter-efficient fine-tuning (PEFT) alternatives have been promoted recently. However, the existing PEFT algorithms typically follow a bottom-up way, where their attention solely relies on the input and lacks the capability of task-guided top-down attention, which provides the task-relevant representation such as the human visual perception system. In this article, we introduce ReFocus, a new PEFT method that adapts the pretrained RGB foundation tracking model to the downstream TIR tracking task through the guidance of high-level task-specific signals in a top-down attention manner. By freezing the entire foundation model and only training query-guided feature selection and top-down blocks, ReFocus achieves state-of-the-art (SOTA) TIR tracking performance while keeping training efficiency. Extensive experiments on five TIR tracking benchmarks demonstrate that ReFocus significantly improves the performance of the foundation tracker. Besides, further ablation studies show the effectiveness and flexible adaptability of the proposed method to lighter foundation models and different tracking frameworks. Compared to FFT and other bottom-up PEFT paradigms, such as head probe, low-rank adaptation (LoRA), and adapter, our method achieves comparable or superior performance with fewer training parameters and reveals the advantage of learning stability.