Search Results (1,700)

Recent advances in deep learning have led to a proliferation of AI services for the general public. Consequently, constructing trustworthy AI systems that operate on personal devices has become a crucial challenge. While on-device processing is critical for privacy-preserving and latency-sensitive applications, conventional deep learning approaches often suffer from instability under quantization and high computational costs. Toward a trustworthy and efficient on-device solution for image processing, we present a hybrid neural filtering framework that combines the representational power of lightweight neural networks with the stability of classical filters. In our framework, the neural network predicts a low-dimensional parameter map that guides the filter’s behavior, effectively decoupling parameter estimation from the final image synthesis. This design enables a truly trustworthy AI system by operating entirely on-device, which eliminates the reliance on servers and significantly reduces computational cost. To ensure quantization robustness, we introduce a basis-decomposed parameterization, a design mathematically proven to bound reconstruction errors. Our network predicts a set of basis maps that are combined via fixed coefficients to form the final guidance. This architecture is intrinsically robust to quantization and supports runtime-adaptive precision without retraining. Experiments on depth map super-resolution validate our approach. Our framework demonstrates exceptional quantization robustness, exhibiting no performance degradation under 8-bit quantization, whereas a baseline suffers a significant 1.56 dB drop. Furthermore, our model’s significantly lower Mean Squared Error highlights its superior stability, providing a practical and mathematically grounded pathway toward trustworthy on-device AI. Full article

Ancient paintings, as invaluable cultural heritage, often suffer from damages like creases, mold, and missing regions. Current restoration methods, while effective for natural images, struggle with the fine-grained control required for ancient paintings’ artistic styles and brushstroke patterns. We propose the Semantic and Sketch-Guided Restoration (SSGR) framework, which uses pixel-level semantic maps to restore missing and mold-affected areas and depth-aware sketch maps to ensure texture continuity in creased regions. The sketch maps are automatically extracted using advanced methods that preserve original brushstroke styles while conveying geometry and semantics. SSGR employs a semantic segmentation network to categorize painting regions and depth-sensitive sketch extraction to guide a diffusion model. To enhance style controllability, we cluster diverse attributes of landscape paintings and incorporate a Semantic-Sketch-Attribute-Normalization (SSAN) block that explores consistent patterns across styles through spatial semantic and attribute-adaptive normalization modules. Evaluated on the CLP-2K dataset, SSGR achieves an mIoU of 53.30%, SSIM of 0.42, and PSNR of 13.11, outperforming state-of-the-art methods. This approach not only preserves historical aesthetics but also advances digital heritage preservation with a tailored, controllable technique for ancient paintings. Full article

This study focuses on the experiences of female Paralympic athletes in Brazil through the lens of feminist disability studies. It is a qualitative study, grounded in the voices of disabled women, positioning them as the foundation of knowledge within a post-structuralist epistemological framework. The research involved in-depth interviews with four Brazilian Female Paralympians, conducted using a predefined interview guide, and the data were analyzed using thematic content analysis. The athletes were between 25 and 34 years of age and had experience competing in international competitions. From their narratives, two distinct yet interconnected categories emerge: (1) Being a disabled woman and (2) Like sportswomen. Disabled women have shown that they navigate an ableist and sexist society as disabled women within the context of sport; at the same time, they embody an empowered and inspirational identity as sportswomen. By exploring these perspectives, this study highlights the need to challenge and redefine societal perceptions and expectations surrounding disability. It provides insights into the experiences and agency of female Paralympic athletes. Full article

To address uncertainties in how threatened coastal China seas fish communities respond to stressors like overfishing and climate change, this study applied Hierarchical Modelling of Species Communities (HMSC) to disentangle the assembly rules shaping these communities, filling a critical gap in understanding their spatiotemporal dynamics. We analyzed data on 384 fish species (1980–2018) and key environmental factors, with variance partitioning revealing that environmental filtering dominated fish distributions (explaining over 99% of variance), far outweighing random effects (0.60%). Among environmental drivers, sea surface temperature (49.00%) and sea surface salinity (33.25%) were the most influential, while seafloor substrate and water depth played secondary roles; notably, fewer species occupied fine sand habitats, and more preferred silt habitats. Residual species associations—indicative of potential biotic interactions—were most frequent within Gobiidae, likely due to this highly diverse taxon’s specialized resource utilization and wide distribution, highlighting that biotic filtering is concentrated and ecologically relevant within this group. This work demonstrates HMSC’s utility in unraveling coastal fish community assembly, providing a robust basis for predicting community changes and guiding biodiversity conservation efforts that support ocean health and dependent human activities. Full article

Accurate prediction of the rate of penetration (ROP) in deep and ultra-deep wells remains a major challenge due to complex downhole conditions and limited real-time data. To address the issues of physical inconsistency and weak generalization in conventional da-ta-driven approaches, this study proposes a mechanism-guided and attention-enhanced deep learning framework. In this framework, drilling physical principles such as energy balance are reformulated into differentiable constraint terms and directly incorporated in-to the loss function of deep neural networks, ensuring that model predictions strictly ad-here to drilling physics. Meanwhile, attention mechanisms are integrated to improve feature selection and temporal modeling: for tree-based models, we investigate their implicit attention to key parameters such as weight on bit (WOB) and torque; for sequential models, we design attention-enhanced architectures (e.g., LSTM and GRU) to capture long-term dependencies among drilling parameters. Validation on 49,284 samples from 11 deep and ultra-deep wells in China (depth range: 1226–8639 m) demonstrates that the synergy between mechanism constraints and attention mechanisms substantially improves ROP prediction accuracy. In blind-well tests, the proposed method achieves a mean absolute percentage error (MAPE) of 9.47% and an R² of 0.93, significantly outperforming traditional methods under complex deep-well conditions. This study provides reliable intelligent decision support for optimizing deep and ultra-deep well drilling operations. By improving prediction accuracy and enabling real-time anomaly detection, it enhances operational safety and efficiency while reducing drilling risks. The proposed approach offers high practical value for field applications and supports the intelligent development of the oil and gas industry. Full article

Background/Objectives: The purpose of the study is to determine and compare the accuracy and efficiency of two dynamic navigation systems (DNS)—Navident (ClaroNav, Canada) and X-Guide (Nobel Biocare, Switzerland)—vs. a free-hand (FH) approach in performing endodontic microsurgery (EMS) on human cadavers. Methods: a total of 119 roots of six cadavers were randomly divided into three groups (Navident/X-Guide/FH). The cadavers’ jaws were scanned pre-operatively with computed tomography. The DICOM data were uploaded and digitally managed with software interfaces for registration, calibration, and virtual planning of EMS. Osteotomy was performed under DNS control and using a dental operating microscope (FH control group). Post-operative scans were taken with same settings as preoperative. Accuracy was then determined by comparing pre- and post-scans of coronal and apical linear, angular deviation, angle, length, and depth of apical resection. Efficiency was determined by measuring the procedural time of osteotomy, apicectomy, retro-cavity preparation, the volume of substance, and cortical bone loss, as well as iatrogenic complications. Outcomes were also evaluated in relation to different operators’ skill levels. Descriptive statistics and inferential analyses were conducted using R software (4.2.1). Results: DNS demonstrated better efficiency in osteotomy and apicectomy, second only to FH in substance and cortical bone loss. Both DNS approaches had similar accuracy. Experts were faster and more accurate than non-experts in FH, apart from resection angle, length and depth, and retro-cavity preparation time, for which comparison was not statistically significant. The Navident and X-guide groups had similar trends in increasing efficiency and accuracy of EMS. All complications in the FH group were performed by non-experts. The X-guide group demonstrated fewer complications than the Navident group. Conclusions: Both DNS appear beneficial for EMS in terms of accuracy and efficacy in comparison with FH, also demonstrating the decreasing gap of skill expertise between experts and novice operators. Through convenient use X-guide diminishes the level of iatrogenic complications compared to Navident. Full article

Stereo matching estimates disparity by finding correspondences between stereo image pairs. Under ill-posed conditions such as geometric differences, radiometric differences, and temporal changes, accurate estimation becomes difficult due to insufficient matching information. In remote sensing imagery, such ill-posed regions are more common because of complex imaging conditions. This problem is particularly pronounced in cross-track satellite stereo images, where existing methods often fail to effectively handle noise due to insufficient features or excessive reliance on prior assumptions. In this work, we propose an ill-posed-aware aggregated satellite stereo matching network, which integrates monocular depth estimation with an ill-posed-guided adaptive aware geometry fusion module to balance local and global features while reducing noise interference. In addition, we design an enhanced mask augmentation strategy during training to simulate occlusions and texture loss in complex scenarios, thereby improving robustness. Experimental results demonstrate that our method outperforms existing state-of-the-art approaches on the US3D dataset, achieving a 5.38% D1-error and 0.958 pixels endpoint error (EPE). In particular, our method shows significant advantages in ill-posed regions. Overall, the proposed network not only exhibits strong feature learning ability but also demonstrates robust generalization in real-world remote sensing applications. Full article

International postgraduate students enrich higher education institutions and host societies, contributing economically, socially, and culturally. However, much less is known about how to improve their academic outcomes compared with their undergraduate counterparts. This study explores the impact of a non-traditional form of learning, a six-week immersive block model underpinned by guided, active learning pedagogy, on the academic success, satisfaction, and experiences of international postgraduate students at an Australian university. A convergent mix-methods design was used. Chi square tests and generalised estimating equations were used to compare the students’ success rates (N = 14,340) and unit satisfaction (N = 4903) in traditional semester and immersive block learning over five years. Qualitative insights were gathered via student focus groups (N = 9). Significant positive changes in success were observed after controlling for gender, age, discipline, and home region, with particularly strong positive effects for male and information technology students. Despite some challenges with depth of learning and placement organisation, focus group participants valued the clear timelines and flexible delivery, reporting that this supported effective time management and study-work–life-balance. Immersive block learning appears to be an effective strategy for transforming the experiences and outcomes of international postgraduate students in higher education. Full article

Autonomous underwater vehicles (AUVs) play a critical role in underwater remote sensing and monitoring applications. This paper addresses the problem of navigating multiple AUVs to perform sweep video sensing of unknown underwater regions over uneven seafloors, where visibility is limited by the conical field of view (FoV) of the onboard cameras and by occlusions caused by terrain. Coverage is formulated as a feasibility objective of achieving a prescribed target fraction while respecting vehicle kinematics, actuation limits, terrain clearance, and inter-vehicle spacing constraints. We propose an online, occlusion-aware trajectory planning algorithm that integrates frontier-based goal selection, safe viewing depth estimation with clearance constraints, and model predictive control (MPC) for trajectory tracking. The algorithm adaptively guides a team of AUVs to preserve line of sight (LoS) visibility, maintain safe separation, and ensure sufficient clearance while progressively expanding coverage. The approach is validated through MATLAB simulations on randomly generated 2.5D seafloor surfaces with varying elevation characteristics. Benchmarking against classical lawnmower baselines demonstrates the effectiveness of the proposed method in achieving occlusion-aware coverage in scenarios where fixed-pattern strategies are insufficient. Full article

Objectives: Venous malformations (VMs) that infiltrate the muscular layer, involve or are closely adjacent to critical nerves or vessels, or are located deep within or very close to major organs in the thoracic or abdominal cavities are challenging to access during sclerotherapy, which we defined as inaccessible VMs. This study proposed an integrated real-time stereotactic MRI-guided sclerotherapy with bleomycin-polidocanol foam (RSMS-BPF) for the treatment of inaccessible VMs, focusing on its clinical feasibility, efficacy, and safety. Methods: A retrospective study was conducted involving patients treated with RSMS-BPF between 2019 and 2021. During the sclerotherapy, the intraoperative magnetic resonance imaging (MRI) was combined with an optical navigation system to guide precise needle placement and track BPF, a foam sclerosant optimized for MRI visibility. Radiological response was assessed by lesion volume, while clinical improvement was evaluated through patients’ description of their symptoms. Rigorous follow-up and documentation of complications were conducted. Results: A total of 42 patients (mean age 23.6 ± 1.6 years; 18 males) were treated in 64 sclerotherapy sessions. The treatment achieved an overall response rate of 89.5%. Imaging analysis revealed an average lesion volume reduction of 59.6%. 57.9% of patients achieved good or excellent radiological responses. After a median follow-up of 12.25 months, 60.53% of patients reported complete or significant relief. Lesion depth did not affect treatment efficacy (p = 0.43). Minor complications included skin hyperpigmentation (5.3%, 2/38) and blisters (2.6%, 1/38), with no major complications observed. Conclusions: RSMS-BPF demonstrated satisfactory efficacy and safety in VMs treatment, particularly for inaccessible VM lesions. It enables authentic real-time dynamic tracking during sclerotherapy, achieving unparalleled precision targeting while minimizing procedural risks. These findings strongly support routine integration of RSMS-BPF as first-line therapy for complex vascular malformations with critical anatomical constraints. Full article

Background: Contralateral lymph node metastasis (CLNM) in oral squamous cell carcinoma (OSCC) represents a major clinical challenge, in patients with a clinically contralateral node-negative neck. Individualized risk stratification is crucial to guide decisions on elective contralateral neck dissection. This study aimed to synthesize existing evidence and apply Bayesian Monte Carlo Simulation (MCS) to estimate CLNM probability across various clinic-pathological scenarios. Methods: A systematic search of PubMed, PubMed Central, and Embase (2000–2024) identified 26 eligible studies. Effect sizes for seven key risk factors—midline-crossing tumours, extranodal extension (ENE), ≥2 ipsilateral lymph nodes, depth of invasion (DOI) >10 mm, perineural invasion and lymphovascular invasion (PNI-LVI), poor differentiation, and floor of mouth subsite—were computed and incorporated into a Bayesian logistic model. Using the No-U-Turn Sampler (NUTS) in RStan, 100,000 virtual patient profiles were simulated to generate posterior probabilities of CLNM. Results: The baseline CLNM risk for lateralized tumours without additional risk factors was 4.2%. Single risk factors increased probability substantially: midline-crossing tumours (31.7%), ENE (27.4%), and ≥2 ipsilateral nodes (24.9%). Combinations of risk factors amplified the risk non-linearly: the presence of a midline-crossing tumour, ENE, and ≥2 ipsilateral nodes yielded a 76.8% CLNM probability, and the presence of all seven risk factors increased it to 93.7%. Risk tiers were classified from minimal (<20%) to very high (>50%) to guide clinical decision-making. Conclusions: This MCS-based model reveals that CLNM risk increases multiplicatively with the presence of various high-risk features. The simulation supports bilateral neck management in high-risk patients and observation in low-risk cases. Prospective validation is needed to integrate this model into routine clinical practice and to guide patient-specific surgical planning. Full article

This paper presents a comprehensive review that provides an in-depth examination of humanoid heads, focusing on their mechanics, perception systems, computational frameworks, and human–robot interaction interfaces. The integration of these elements is crucial for developing advanced human–robot interfaces that enhance user interaction and experience. Key topics include the principles of context, functionality, and appearance that guide the design of humanoid heads. This review delves into the different aspects of human–robot interaction, emphasizing the role of artificial intelligence and large language models in improving these interactions. Technical challenges such as the uncanny valley phenomenon, facial expression synthesis, and multi-sensory integration are further explored. This paper identifies future research directions and underscores the importance of interdisciplinary collaboration in overcoming current limitations and advancing the field of humanoid head technology. Full article

Bridges, as critical transportation infrastructure, are highly vulnerable to aerodynamic forces, particularly vortex-induced vibrations (VIV), which severely compromise their structural integrity and operational safety. These low-frequency, high-amplitude vibrations are a primary challenge to serviceability and fatigue life. Ensuring the resilience of these structures demands advanced understanding and robust mitigation strategies. This paper comprehensively addresses the multifaceted challenges of bridge aerodynamics, presenting an in-depth analysis of contemporary testing methodologies and innovative solutions. We critically examine traditional wind tunnel modeling, elucidating its advantages and inherent limitations, such as scale effects, Reynolds number dependence, and boundary interference, which can lead to inaccurate predictions of aerodynamic forces and vibration amplitudes. This scale discrepancy is critical, as demonstrated by peak pressure coefficients being underestimated by up to 64% in smaller-scale wind tunnel environments compared to high-Reynolds-number open-jet testing. To overcome these challenges, the paper details the efficacy of open-jet testing at facilities like the Windstorm Impact, Science, and Engineering (WISE) Laboratory, demonstrating its superior capability in replicating realistic atmospheric boundary layer flow conditions and enabling larger-scale, high-Reynolds-number testing for more accurate insights into bridge behavior under dynamic wind loads. Furthermore, we explore the design principles and applications of various aerodynamic mitigation devices, including handrails, windshields, guide vanes, and spoilers, which are essential for altering airflow patterns and suppressing vortex-induced vibrations. The paper critically investigates the innovative integration of green energy solutions, specifically solar panels, with bridge structures. This study presents the application of solar panel arrangements to provide both renewable energy production and verifiable aerodynamic mitigation. This strategic incorporation is shown not only to harness renewable energy but also to actively improve aerodynamic performance and mitigate wind-induced vibrations, thereby fostering both bridge safety and sustainable infrastructure development. Unlike previous studies focusing primarily on wind loads on PV arrays, this work demonstrates how the specific geometric integration of solar panels can serve as an active aerodynamic mitigation device for bridge decks. This dual functionality—harnessing renewable energy while simultaneously serving as a passive geometric countermeasure to vortex-induced vibrations—marks a novel advancement over single-purpose mitigation technologies. Through this interdisciplinary approach, the paper seeks to advance bridge engineering towards more resilient, efficient, and environmentally responsible solutions. Full article

Men who have sex with men (MSM) are part of the key populations (KPs) that are susceptible to Human Immunodeficiency Virus (HIV) acquisition, transmission, and disruptions to access HIV services. This study aimed to explore and describe the HIV interventions implemented among MSM during the COVID-19 lockdown in the Capricorn District of Limpopo Province in South Africa. This study followed an exploratory qualitative study design with a purposive sample of 16 men who identified as MSM. Data were collected through in-depth interviews which were tape-recorded, transcribed verbatim, and captured on Atlas.ti. Patterns from the created codes were formulated into themes/sub-themes guided by the Consolidated Framework for Implementation Research (CFIR). The MSM had a mean age of 27.9 years. Various constructs and domains of the CFIR showed that during the COVID-19 lockdown, MSM encountered a disruption of HIV services. The response time was elongated, thus increasing the complexity of interventions. Interventions were implemented to ensure continuity of services. These included tailoring the programme through a door-to-door strategy and offering resources to healthcare facilities to enable MSM to access antiretroviral therapy (ART) and related services. The relative advantage of the programme is that it prioritises MSM. Future research should look into COVID-19’s impact on a bigger scale. Full article

Accurate depth image inpainting in complex urban environments remains a critical challenge due to occlusions, reflections, and sensor limitations, which often result in significant data loss. We propose a hybrid deep learning framework that explicitly combines local and global modelling through Convolutional Neural Networks (CNNs) and Transformer modules. The model employs a multi-branch parallel architecture, where the CNN branch captures fine-grained local textures and edges, while the Transformer branch models global semantic structures and long-range dependencies. We introduce an optimized attention mechanism, Agent Attention, which differs from existing efficient/linear attention methods by using learnable proxy tokens tailored for urban scene categories (e.g., façades, sky, ground). A content-guided dynamic fusion module adaptively combines multi-scale features to enhance structural alignment and texture recovery. The frame-work is trained with a composite loss function incorporating pixel accuracy, perceptual similarity, adversarial realism, and structural consistency. Extensive experiments on the Paris StreetView dataset demonstrate that the proposed method achieves state-of-the-art performance, outperforming existing approaches in PSNR, SSIM, and LPIPS metrics. The study highlights the potential of multi-scale modeling for urban depth inpainting and discusses challenges in real-world deployment, ethical considerations, and future directions for multimodal integration. Full article