Search Results (1,182)

Human Action Recognition has seen significant advances through transformer-based architectures, yet achieving a nuanced understanding often requires fusing multiple data modalities. Standard models relying solely on RGB video can struggle with actions defined by subtle motion cues rather than appearance. This paper introduces TransMODAL, a novel dual-stream transformer that synergistically fuses spatiotemporal appearance features from a pre-trained VideoMAE(Video Masked AutoEncoders) backbone with explicit skeletal kinematics from a state-of-the-art pose estimation pipeline (RT-DETR(Real-Time DEtection Transformer) + ViTPose++). We propose two key architectural innovations to enable effective and efficient fusion: a CoAttentionFusion module that facilitates deep, iterative cross-modal feature exchange between the RGB and pose streams, and an efficient AdaptiveSelector mechanism that dynamically prunes less informative spatiotemporal tokens to reduce computational overhead. Evaluated on three challenging benchmarks, TransMODAL demonstrates robust generalization, achieving accuracies of 98.5% on KTH, 96.9% on UCF101, and 84.2% on HMDB51. These results significantly outperform a strong VideoMAE-only baseline and are competitive with state-of-the-art methods, demonstrating the profound impact of explicit pose guidance. TransMODAL presents a powerful and efficient paradigm for composing pre-trained foundation models to tackle complex video understanding tasks by providing a fully reproducible implementation and strong benchmark results. Full article

Single remote sensing image dehazing aims to eliminate atmospheric scattering effects without auxiliary information. It serves as a crucial preprocessing step for enhancing the performance of downstream tasks in remote sensing images. Conventional approaches often struggle to balance haze removal and detail restoration under non-uniform haze distributions. To address this issue, we propose a Dual-domain Feature Fusion Network (DFFNet) for remote sensing image dehazing. DFFNet consists of two specialized units: the Frequency Restore Unit (FRU) and the Context Extract Unit (CEU). As haze primarily manifests as low-frequency energy in the frequency domain, the FRU effectively suppresses haze across the entire image by adaptively modulating low-frequency amplitudes. Meanwhile, to reconstruct details attenuated due to dense haze occlusion, we introduce the CEU. This unit extracts multi-scale spatial features to capture contextual information, providing structural guidance for detail reconstruction. Furthermore, we introduce the Dual-Domain Feature Fusion Module (DDFFM) to establish dependencies between features from FRU and CEU via a designed attention mechanism. This leverages spatial contextual information to guide detail reconstruction during frequency domain haze removal. Experiments on the StateHaze1k, RICE and RRSHID datasets demonstrate that DFFNet achieves competitive performance in both visual quality and quantitative metrics. Full article

Accurately predicting stock prices is crucial for investment and risk management, but the non-stationarity of the financial market and the complex correlations among stocks pose challenges to traditional models (ARIMA, LSTM, XGBoost), resulting in difficulties in effectively capturing dynamic patterns and limited prediction accuracy. To this end, this paper proposes the Financial Spatio-Temporal Graph Attention Network (FSTGAT), with the following core innovations: temporal modelling through gated causal convolution to avoid future information leakage and capture long- and short-term fluctuations; enhanced spatial correlation learning by adopting the Dynamic Graph Attention Mechanism (GATv2) that incorporates industry information; designing the Multiple-Input-Multiple-Output (MIMO) architecture of industry grouping for the simultaneous learning of intra-group synergistic and inter-group influence; symmetrically fusing spatio-temporal modules to construct a hierarchical feature extraction framework. Experiments in the commercial banking and metals sectors of the New York Stock Exchange (NYSE) show that FSTGAT significantly outperforms the benchmark model, especially in high-volatility scenarios, where the prediction error is reduced by 45–69%, and can accurately capture price turning points. This study confirms the potential of graph neural networks to model the structure of financial interconnections, providing an effective tool for stock forecasting in non-stationary markets, and its forecasting accuracy and industry correlation capturing ability can support portfolio optimization, risk management improvement and supply chain decision guidance. Full article

Understanding the mechanisms governing forest community assembly across different growth stages is essential for revealing succession dynamics and guiding forest restoration. While much attention has been given to overstory trees, the understory regeneration layer, critical for forest succession, remains less explored, particularly regarding its stage-specific survival strategies and assembly processes. This study investigates the natural regeneration of Quercus variabilis forests in northern China, focusing on the transition from early to later growth stages. Our objectives were to (1) identify the phylogenetic and functional structures of regeneration communities at early and later stages, (2) explore their responses to environmental gradients, and (3) assess the roles of deterministic and stochastic processes in shaping community assembly. We integrated phylogenetic structure, functional traits, and environmental gradients to examine natural regeneration communities. The results revealed clear stage-dependent patterns: communities exhibited random phylogenetic and functional structures in the early growth stage, suggesting a dominant role of stochastic processes during early recruitment. In contrast, communities showed phylogenetic clustering and functional overdispersion in later growth stages, indicating the increasing influence of environmental filtering and interspecific competition as individuals developed. Generalized Dissimilarity Modeling (GDM) further revealed that dispersal limitation and pH were key predictors of phylogenetic β-diversity in the later growth stage, while total phosphorus drove functional β-diversity in the later growth stage. No significant predictors were found for β-diversity in the early stage. These findings highlight the shift from stochastic to deterministic processes during forest regeneration, emphasizing the stage-dependent nature of assembly mechanisms. Our study elucidates the stage-specific assembly rules of Q. variabilis forests and offers theoretical guidance for stage-targeted interventions in forest management to promote positive succession. Full article

Background: This study stems from the perceived need to update skills and training in the process of educating healthcare professionals in light of the needs of individuals and their families. Objectives: This study aimed to assess the prevailing attitudes toward death and hospice care among medical students in China, providing a foundation for implementing hospice care and death education within these institutions. Methods: We conducted an online survey questionnaire with 568 medical students. Results: The results indicate that the overall attitude toward death was more inclined to accepting death neutrally. Gender, place of origin, educational background, willingness to care for terminally ill patients, experience in caring for terminally ill patients, and more will affect the attitude toward death of medical students. Compared to their rural counterparts, medical students in urban areas are more likely to view death as neutral. Instead of reducing fear, death and hospice education made people more likely to avoid situations. All five dimensions of death attitude exhibit a substantial positive connection with attitudes toward hospice care. In comparison to earlier research, medical students exhibit increasingly favorable attitudes regarding hospice care, and their overall perspective on death remains one of natural approval, suggesting that the integration of death and hospice care teaching is progressing effectively in mainland China. Conclusions: Simultaneously, it was discovered that numerous deficiencies required enhancement, including the need for timely feedback and optimization in hospice care instruction and death education, as well as insufficient attention and educational guidance regarding the individual differences and psychological conditions of medical personnel in the future. Full article

The shift toward experience-oriented travel has positioned food as a central driver for attracting visitors to sustainable destinations, directly supporting United Nations Sustainable Development Goal (SDG)11 (resilient cities) and SDG 12 (responsible consumption). While prior research has predominantly emphasised marketing outcomes, the role of bodily experiences in shaping gastronomic tourism has received less attention. This study explores how sensory elements (sight, sound, smell, taste, and touch) and non-sensory elements (including cultural meaning and service quality) jointly influence food-related travel experiences. Twenty-five self-identified food travellers were interviewed in a United Nations Educational, Scientific and Cultural Organization (UNESCO) City of Gastronomy, and their narratives were analysed using a three-stage grounded theory approach in NVivo 12. The resulting model identifies four interrelated dimensions: (1) embodied experience, grounded in culinary memories and shared cultural narratives; (2) sensory stimulation arising from food and its surroundings; (3) situated embodiment, shaped by location, timing, and social interaction; and (4) environmental perception, encompassing food presentation, facility quality, cleanliness, and pricing fairness. These dimensions interact to enhance overall experience quality. By integrating an embodied perspective with a sustainability focus, this study advances tourism experience research and offers practical guidance for designing multisensory dining environments, fostering environmentally responsible visitor behaviour, and ensuring a balanced relationship between price and perceived value. Full article

Nanomaterials are materials in which at least one dimension in three-dimensional space is at the nanoscale. In recent years, nano-alumina has attracted much attention due to its large specific surface area and pore volume, as well as novel optical, magnetic, electronic, and catalytic properties. This review summarizes the preparation methods of nano-alumina based on the basic phases and properties of alumina materials, focusing on one-dimensional, two-dimensional, and three-dimensional nano-alumina preparation methods, which can provide some theoretical guidance for the subsequent development of efficient nano-alumina materials. Finally, the application of nano-alumina materials in catalysis is reviewed, and some suggestions are provided for improving the use of nano-alumina in the catalysis field. Full article

The remarkable success of autoregressive Large Language Models (LLMs) is predicated on the causal attention mechanism, which enforces a static and rigid form of informational asymmetry by permitting each token to attend only to its predecessors. While effective for sequential generation, this hard-coded unidirectional constraint fails to capture the more complex, dynamic, and nonlinear dependencies inherent in sophisticated reasoning, logical inference, and discourse. In this paper, we challenge this paradigm by introducing Dynamic Asymmetric Attention (DAA), a novel mechanism that replaces the static causal mask with a learnable context-aware guidance module. DAA dynamically generates a continuous-valued attention bias for each query–key pair, effectively learning a “soft” information flow policy that guides rather than merely restricts the model’s focus. Trained end-to-end, our DAA-augmented models demonstrate significant performance gains on a suite of benchmarks, including improvements in perplexity on language modeling and notable accuracy boosts on complex reasoning tasks such as code generation (HumanEval) and mathematical problem-solving (GSM8k). Crucially, DAA provides a new lens for model interpretability. By visualizing the learned asymmetric attention patterns, it is possible to uncover the implicit information flow graphs that the model constructs during inference. These visualizations reveal how the model dynamically prioritizes evidence and forges directed logical links in chain-of-thought reasoning, making its decision-making process more transparent. Our work demonstrates that transitioning from a static hard-wired asymmetry to a learned and dynamic one not only enhances model performance but also paves the way for a new class of more capable and profoundly more explainable LLMs. Full article

While blockchain technology (BT) has gained attention in the construction industry, limited research has focused on its application in sustainable construction procurement management (SCPM). Addressing this gap, the present study investigates the key drivers influencing BT adoption in SCPM using a hybrid methodological approach. This study includes a systematic review of academic and grey literature, expert consultations, and quantitative analysis using advanced fuzzy-based algorithms, k-means clustering, and social network analysis (SNA). Data were collected through an online survey distributed to professionals experienced in SCPM and blockchain implementation. The Fuzzy DEMATEL results identify “high quality”, “decentralization and data security”, and “cost of the overall project” as the most critical drivers. Meanwhile, SNA highlights “stability of the system”, “overall performance of the project”, and “customer satisfaction” as the most influential nodes within the network. These insights provide actionable guidance for industry stakeholders aiming to advance SCPM through blockchain integration and contribute to theoretical advancements by proposing novel analytical frameworks. Full article

Spatial environments influence users’ behavioral patterns and psychological perceptions, affecting health outcomes—a professional consensus in architecture, particularly within healthy buildings. Growing attention to spatial design’s health benefits has rapidly increased quantitative research. Relationships between spatial elements (e.g., green spaces, water features, facilities) and health indicators (e.g., emotional state, mental health, physical activity) are increasingly clear. Due to collective behavior patterns on campuses, the space–health relationship is particularly pronounced. This paper examines campus open spaces via meta-analysis to explore spatial elements’ relative influence on health outcomes. After a chronological review of qualitative research, it cross-sectionally extracts quantitative data. The independent variable (“campus open space”) is categorized into natural landscapes, service facilities, and built environment (design organization). The dependent variable (“health”) is subdivided into physical health, mental health, and positive social adaptation. The main conclusions of the study are as follows: Campus open spaces significantly impact student health, with the built environment exerting the strongest influence. Combining landscape/facility elements with spatial guidance yields more significant results. Furthermore, based on the calculated impact factor data for each element, this study has developed an evaluation scale that could serve as an empirical foundation for future assessments of campus health benefits, thereby guiding health-oriented campus spatial design. Full article

PCBs play a critical role in electronic manufacturing, and accurate defect detection is essential for ensuring product quality and reliability. However, PCB defects are often small, irregularly shaped, and embedded in complex textures, making them difficult to detect using traditional methods. In this paper, we propose CM-UNetv2, a semantic segmentation network designed to address these challenges through three architectural modules incorporating four key innovations. First, a Parallelized Patch-Aware Attention (PPA) module is incorporated into the encoder to enhance multi-scale feature representation through a multi-branch attention mechanism combining local, global, and serial convolutions. Second, we propose a Dual-Stream Skip Guidance (DSSG) module that decouples semantic refinement from spatial information preservation via two separate skip pathways, enabling finer detail retention. Third, we design a decoder module called Frequency-domain Guided Context Mamba (FGCMamba), which integrates two novel mechanisms: a Spatial Guidance Cross-Attention (SGCA) mechanism to enhance the alignment of spatial and semantic features, and a Frequency-domain Self-Attention Solver (FSAS) to compute global attention efficiently in the frequency domain, improving boundary restoration and reducing computational overhead. Experiments on the MeiweiPCB and KWSD2 datasets demonstrate that the CM-UNetv2 achieves state-of-the-art performance in small object detection, boundary accuracy, and overall segmentation robustness. Full article

The environmental pollution problem caused by the construction industry has been paid attention to by scholars. However, few existing studies on supply chain management explore the interplay between information-sharing strategies and green-effort strategies in a green building materials supply chain. This study explores green building materials design and information-sharing dynamics in a supply chain consisting of a common building enterprise and two competing building materials manufacturers. The building enterprise decides whether to share demand information with manufacturers, who then determine product greenness, while the building enterprise determines the retail price. The findings reveal that information sharing has dual effects on manufacturers’ profitability, depending on competitive dynamics and demand sensitivity to building materials greenness. Additionally, the interplay between information sharing and green design strategies highlights the importance of aligning product design decisions with optimal information-sharing practices. While information sharing consistently improves environmental performance in a bilateral monopoly system where a single manufacturer provides building materials to a single building enterprise, it can induce adverse environmental outcomes in competitive scenarios. These results provide actionable guidance for developing green supply chain strategies that balance economic and environmental goals. Full article

(1) Deficits in adaptive behavior significantly hinder individuals with intellectual disabilities from performing essential daily tasks and participating in community life. Although virtual reality shows promise for supporting adaptive behavior in this population, systematic reviews on this topic remain scarce. (2) Methods: Twenty-five experimental studies from the databases Web of Science, PubMed, Scopus, and ERIC, published between 2005 and 2024, were analyzed in the context of a systematic review. (3) Results: The studies revealed a significant surge in research on VR interventions for adaptive behavior in individuals with intellectual disabilities, particularly after 2021. The most frequently applied domain was practical skills, while social and conceptual skills received relatively less attention. Most studies employed high-immersion head-mounted displays as the primary technology type and adopted controller-based unimodal interaction as the dominant interaction mode. Pedagogical strategies such as ABA, structured teaching, and contextual learning are favored in interventions. (4) Conclusions: VR interventions have been increasingly applied to support adaptive behavior development in this population. However, further exploration is needed to tailor VR designs to better accommodate the individual differences and specific needs. This review synthesizes current evidence, identifies key trends and limitations, and offers guidance for future research. Full article

Due to human activities and the invasion of Spartina alterniflora, the population of Scirpus mariqueter (S. mariqueter) in the Yangtze River Estuary has gradually declined. To address this issue, numerous restoration efforts have been undertaken. To investigate the changes and influencing factors of soil bacterial communities during the restoration of S. mariqueter wetlands, we selected S. mariqueter populations as the research focus and divided the samples into two years, S1 and S2. High-throughput sequencing technology was employed for observation and analysis. The results revealed that from S1 to S2, soil bacterial diversity in the S. mariqueter wetland increased significantly and displayed clear seasonal patterns. The dominant bacterial phyla included Proteobacteria, Bacteroidota, Firmicutes, and Acidobacteriota. Among these, Proteobacteria had the highest relative abundance, while Acidobacteriota showed the most pronounced increase, surpassing Bacteroidota and Firmicutes to become the second most abundant group. Redundancy analysis (RDA) indicated that soil organic matter and electrical conductivity were the key factors influencing the composition and diversity of the soil bacterial community, with Acidobacteriota playing a dominant role during wetland restoration. In conclusion, during the ecological restoration of the S. mariqueter wetlands, attention should be given to environmental factors such as soil organic matter and electrical conductivity, while the regulatory role of Acidobacteriota in wetland soils should not be overlooked. This study provides a microscopic perspective on the interactions between microbial diversity and ecosystem functions in coastal wetlands, offering valuable guidance for the ecological restoration of degraded wetlands. Full article

Aqueous redox flow batteries (ARFBs) have attracted significant attention in the field of electrochemical energy storage due to their high intrinsic safety, low cost, and flexible system configuration. However, the advancement of this technology is still hindered by several critical challenges, including capacity decay, structural optimization, and the design and application of key materials as well as their performance within battery systems. Addressing these issues requires systematic theoretical foundations and scientific guidance. Numerical modeling has emerged as a powerful tool for investigating the complex physical and electrochemical processes within flow batteries across multiple spatial and temporal scales. It also enables predictive performance analysis and cost-effective optimization at both the component and system levels, thus accelerating research and development. This review provides a comprehensive overview of recent progress in the modeling of ARFBs. Taking the all-vanadium redox flow battery as a representative example, we summarize the key multiphysics phenomena involved and introduce corresponding multi-scale modeling strategies. Furthermore, specific modeling considerations are discussed for phase-change ARFBs, such as zinc-based ones involving solid–liquid phase transition, and hydrogen–bromine systems characterized by gas–liquid two-phase flow, highlighting their distinctive features compared to vanadium systems. Finally, this paper explores the major challenges and potential opportunities in the modeling of representative ARFB systems, aiming to provide theoretical guidance and technical support for the continued development and practical application of ARFB technology. Full article