Search Results (1,361)

Affective Image Stylization (AIS) converts an emotional intent into executable artistic visual styles. Existing methods are often limited to discrete label settings and provide limited interpretability of how target emotions are realized. We propose LLaVA-Emo, an interpretable AIS framework built on multimodal Chain-of-Thought (CoT) reasoning. Our method decouples generation into two structured outputs: <reasoning> provides visual–affective causal explanations grounded in the input image evidence, and <style_prompt> expresses actionable, renderer-ready style instructions that directly condition a frozen diffusion renderer. We constructed a training set by screening ArtEmis’ sentiment interpretations and fine-tune LLaVA-1.5-7B with LoRA, where SFT mainly supervises the structured intermediate <reasoning> (and output format), while the true executability of <style_prompt> is enforced by our DPO stage via render-and-reward feedback. The rendering stage remains training-free, and we further apply DPO for preference optimization to align candidate outputs with both emotion fidelity and instruction executability. Experiments on the EmoEdit inference set demonstrate that LLaVA-Emo improves emotion alignment while providing stronger process interpretability. Full article

Urban low-altitude unmanned aerial vehicle (UAV) logistics networks face critical operational bottlenecks due to complex three-dimensional spatial blockages, continuous communication diffraction, and severe vulnerability to information-layer threats such as Automatic Dependent Surveillance—Broadcast (ADS-B) signal anomalies. To address these interconnected challenges, this paper proposes an event-driven, cloud–edge collaborative digital twin framework to guarantee continuous multi-link communication and flight safety. The architecture operates through a dual-tier “Teacher–Student” paradigm. Under secure conditions, a cloud digital twin acts as a high-capacity “Teacher,” employing Density-Based Spatial Clustering of Applications with Noise (DBSCAN) to partition heterogeneous user topologies. It then utilizes an energy-guided stochastic diffusion sampling (EGSDS) method to refine initial macroscopic routing, generating precise, outage-free global trajectories by systematically minimizing non-line-of-sight (NLoS) observation penalties and kinematic regularization costs. To counteract signal anomalies, a distributed Time Difference of Arrival (TDOA) anchor network continuously validates UAV coordinate integrity. If a threshold is breached, control authority is instantly transferred to the UAV’s edge digital twin. This resource-constrained edge tier relies on a localized “Student” network trained via progressive distillation. By compressing the computationally heavy iterative diffusion process into a rapid one-step inference model, the UAV autonomously generates a secure, short-range emergency path that strictly adheres to minimum communication thresholds. Once interference clears, the cloud seamlessly regains control to complete the logistics mission. Experimental results demonstrate that the proposed scheme significantly outperforms conventional heuristic routing methods in cloud-based scenarios. Furthermore, the edge-based distillation mechanism substantially improves the overall trajectory survival rate under signal anomalies, ensuring resilient and continuous logistics operations. Full article

Head-mounted eye-tracking systems play a critical role in virtual reality, human–computer interaction, and clinical applications, yet achieving both high angular accuracy and precise 3D gaze position estimation with low-cost hardware remains challenging. This paper proposes a lightweight, training-free geometric 3D gaze tracking framework for binocular 3D gaze tracking using consumer-grade hardware, which leverages stereo geometric triangulation and a simplified physiological eye model to achieve robust 3D gaze estimation, requiring only standard infrared cameras and dichroic mirrors without additional specialized hardware. The method was evaluated in controlled indoor conditions with 30 participants, where it achieved an angular error ranging from 1.1° to 2.82° and a 3D gaze position error below 13.24 mm. Compared to two state-of-the-art academic non-deep-learning methods, the proposed framework delivers competitive angular accuracy while significantly reducing 3D position error, outperforming the baselines by 34% to 56% in depth estimation precision. These results demonstrates that the proposed geometric framework is a practical and effective solution for high-precision 3D gaze tracking on low-cost hardware, suitable for both research and consumer applications. Full article

Reliable monitoring of spontaneous affect is essential in biomedical sensing, where involuntary facial signals serve as objective indicators of physiological states. Micro expression recognition (MER) is particularly challenging due to the sub-second, low amplitude nature of these signals. Many existing MER methods rely on apex (peak) frame detection, making them sensitive to temporal localization errors and difficult to deploy in unconstrained settings. To address this, we propose an apex-free framework that analyzes facial dynamics by structuring motion-magnified features along a newly introduced magnification intensity axis. By applying Eulerian motion magnification across multiple discrete levels and collapsing the sequences into single accumulation images, we generate a multi-level representation of subtle facial dynamics without requiring frame-level annotations. A lightweight shared temporal mixer (STM) is employed to analyze the dynamic evolution of motion across the magnification intensity axis. Subsequently, a dual-branch convolutional neural network (CNN), processing low- and high-amplification regimes respectively, integrates a convolutional block attention module (CBAM) to capture subtle facial motion while effectively filtering out irrelevant noise. Our model is highly efficient, requiring only 0.94 M parameters and 262 MFLOPs, which is significantly lower than the computational demands of standard backbones such as ResNet18 or VGG16. To ensure the model generalizes to new individuals, we evaluated it by testing on subjects whose data was entirely excluded from the training process. Under this rigorous setup, the proposed method achieves approximately 80% and 70% accuracy on the CASME II and SMIC datasets respectively, showing performance comparable to, or in some cases, slightly above current state-of-the-art methods. Considering both the competitive accuracy and high computational efficiency, the proposed framework holds significant potential for practical integration into real-time affect monitoring systems, particularly within biomedical applications. Full article

Automated analysis of peripheral nerve ultrastructure is bottlenecked by heterogeneous electron microscopy (EM) datasets, where varying staining protocols and resolutions create domain shifts that confound deep learning. To address this, we developed a generalized segmentation pipeline. Using a custom pre-processing workflow (CLAHE and noise suppression) integrated into ZEISS Arivis Pro, we standardized inputs across three disparate domains: traditional osmium-based Palade, lanthanide-based “green” Uranyl-free method, and low-resolution Ellisman preparations. A U-Net trained on a highly constrained 15-image composite dataset achieved peak internal Intersection over Union (IoU) scores >0.95 for myelin and Schwann cells. Crucially, during open-world, zero-shot inference on an expanded independent testing cohort (N = 40), the model sustained robust Dice Similarity Coefficients of 0.854 for myelin and 0.597 for mitochondria. This demonstrates that integrating classical image standardization with deep learning effectively mitigates EM domain gaps, enabling comprehensive 3D multi-organelle reconstructions from challenging data. To ensure transparency and community utility, the pre-trained models and standardization scripts are provided in a public, open-access repository. Ultimately, this pipeline supports the transition to sustainable, non-toxic EM protocols and provides a robust pathway for unlocking historical clinical archives for automated organellomics. Full article

Fingerprint mosaicking—the process of combining multiple fingerprint impressions into a single master fingerprint—is an essential step in modern biometric systems, but it is prone to errors that can significantly degrade image quality. This paper proposes a deep learning-based approach to detect and score hard mosaicking artifacts in fingerprint images. Our method uses a self-supervised learning framework to train a segmentation model on large-scale unlabeled fingerprint data, eliminating the need for manual artifact annotation. The proposed model effectively identifies mosaicking errors, achieving high segmentation performance across multiple fingerprint modalities—contactless, rolled, and pressed—and proves robust to different data sources. We also introduce a mosaicking artifact score that quantifies the severity of detected errors and enables automated evaluation of fingerprint images at scale. Training and evaluation rely on synthetic artifacts, we therefore provide a qualitative comparison to real stitching failures and discuss the limits of this validation strategy in detail. By addressing the previously underexplored problem of reference-free hard-artifact detection in fingerprints, our work contributes to improving the accuracy and reliability of fingerprint-based biometric systems. Full article

Assessing motor symptoms in Parkinson’s disease (PD) is challenging due to the progressive evolution of the condition and the variability of symptoms, which are not fully captured by periodic clinical visits. In this context, wearable sensors and machine learning (ML) have emerged as a viable path toward objective and continuous monitoring, although achieving robust generalization to free-living conditions remains a challenge. This work explores the egg-beating task, a simple everyday activity, as a digital approach for PD motor assessment using smartwatch-based inertial measurements and ML techniques. Twenty-two individuals with PD and sixteen healthy controls (HC) completed a one-minute egg-beating task while wearing a smartwatch equipped with tri-axial accelerometer and gyroscope sensors. Data were recorded both under supervised clinical conditions and during unsupervised home sessions. Time- and frequency-domain features were extracted from the inertial signals, and models trained exclusively on supervised recordings were then tested on supervised, unsupervised, and combined data. PD participants showed systematically lower movement amplitude, slower oscillation frequency, and a progressive drop in signal energy over the course of the task, all of which align with the characteristic features of bradykinesia. The support vector machine achieved the best overall performance, reaching 90% accuracy in distinguishing PD from healthy controls under supervised conditions, with a reduction of less than 4% when applied to unsupervised data. These results support the egg-beating task as a practical and ecologically valid method for real-world motor assessment, with potential for future use in remote monitoring and longitudinal assessment. Full article

The main aim of the airborne early warning (AEW) system is to search the potential targets in a large surveillance area. The underlying assumption is that the desired target signals only exist in a few range cells for space-time adaptive processing (STAP), i.e., targets (with certain look direction and Doppler) are sparsely distributed in the entire range cells and most of the range cells are target-free. By utilizing the sparsity of the target distribution, we propose a new STAP method by minimizing the l1-norm of the output magnitude. Unlike conventional STAP methods, which exclude the cell under test from the training samples to avoid target self-nulling, our method processes the cell under test (CUT) and the training samples simultaneously without sample selection. Moreover, to achieve robustness against target steering vector mismatch, we constrain the l1-modulus of the response of any steering vector within a rhombus uncertainty set to exceed unity. Additionally, based on a new definition of the l1-norm of a complex-valued vector, the original nonlinear programming problem can be transformed into a linear programming problem. On the other hand, unlike the slide window processor (SWP) whose weights need to be updated for each range cell, the adaptive weight of our method for a block of samples requires no updating. Consequently, the computational complexity of the proposed method is much lower than that of conventional STAP methods. Finally, since the CUT is used to compute the STAP weights, our method can also suppress the discrete interference. The robustness, computational effectiveness and superiority of the proposed STAP method are verified based on simulated data and the MCARM data. Full article

Text-to-Image Person Re-Identification (TI-ReID) aims to retrieve target pedestrians from large-scale image galleries using natural language descriptions. Despite recent progress achieved by dual-tower architectures based on vision-language pre-training, these methods remain susceptible to semantic misalignment and noise induced by occlusions, background clutter, and fine-grained attribute distractions. To mitigate these issues, we propose a Global Collaborative Discriminative Denoising Network (GCDD), a dual-tower fine-tuning framework built upon a CLIP visual encoder and a BERT text encoder. Specifically, GCDD introduces three complementary branches for robust feature enhancement. First, Discriminative Token Selection (DTS) performs adaptive hard filtering to suppress low-informative tokens. Second, Global-Guided Feature Adaptation (GFA) leverages modality-specific global semantics to recalibrate local features. Third, Query-Driven Aggregation (QDA) constructs more discriminative global representations via attentive pooling, where the backbone global feature serves as the query. The outputs of the three branches are fused through a parameter-free averaging strategy to produce the final representation. Extensive experiments on three standard TI-ReID benchmarks demonstrate that GCDD achieves strong competitive performance, validating the effectiveness of the proposed feature enhancement framework. Full article

US egg production is undergoing a transition to cage-free (CF) housing systems. This transition has increased the need for automated monitoring tools to support welfare management and reduce production costs. While CF houses allow hens to perform natural behaviors such as dust bathing and foraging, a persistent challenge is severe feather pecking. Pecking block enrichment is used as a managemental approach to control severe feather pecking. However, manual quantification of such behavior is subjective and labor-intensive. This study evaluated the performance of small and large variants of both YOLOv10 and YOLO11 models for automatic detection of enrichment block pecking behavior in CF research environment. A total of 1061 color images were used to train and evaluate the models using 70:20:10 split for training, validation, and testing. Performance was assessed using precision, recall, mean average precision at 50% intersection over union (mAP50), confusion matrices, and F1–confidence curve. All models demonstrated robust performance, with precision, recall and mAP50 values greater than 0.94. YOLO11l achieved the highest precision with 0.969 and mAP50 with 0.988, while YOLOv10s achieved the highest recall of 0.962. Evaluation on test datasets showed robust generalization capability of the model, with high confidence detections. Overall, the findings show that YOLO models provide a consistent, objective, and scalable method for automatic quantification of pecking enrichment block related pecking behavior in a CF system. It offers potential as an automated monitoring tool for poultry researchers and may support future development of tools for commercial CF system. Full article

Background: Accurate segmentation of white blood cells (WBCs) in peripheral blood smear images is a fundamental step in computational hematology, enabling downstream tasks such as classification, morphological assessment, and quantitative analysis. However, reliable segmentation remains challenging due to staining variability, complex cellular morphology, overlapping structures, and limited availability of high-quality annotations. Aim and Methods: The aim of this study is to develop a robust and fully automated segmentation framework for white blood cells (WBCs) in microscopic blood smear images, providing a reliable foundation for subsequent computational analysis. We propose TransNet–SAM2, a hybrid deep learning architecture that integrates hierarchical transformer-based feature extraction with a foundation-model-based decoder for prompt-free segmentation. Specifically, a Swin Transformer backbone is employed to capture multi-scale contextual representations, which are subsequently aligned and fused through a feature adaptation module. The fused features are directly injected into the SAM2 mask decoder, replacing conventional prompt-based conditioning and enabling fully automatic segmentation. In addition, a weakly supervised self-training strategy is incorporated to utilize partially annotated data, improving model generalization while reducing annotation requirements. The proposed framework is evaluated using a clinically curated dataset from Dicle University, the publicly available Raabin-WBC dataset, and an additional external leukemic blast validation dataset (ALL-IDB) to assess robustness under both routine and atypical hematological conditions. Results: TransNet-SAM2 achieved a Dice coefficient of 0.95 ± 0.01 and IoU of 0.90 on internal testing, significantly outperforming U-Net (0.89), Mask R-CNN (0.90), and SAM2 (0.92) (p < 0.05). In cross-dataset evaluation (Dicle training, Raabin-WBC testing), the framework maintained strong performance (Dice: 0.91, IoU: 0.84), demonstrating robustness to domain shifts. Ablation studies confirmed each component’s contribution, with the full model improving Dice by 6% over a CNN baseline. Qualitative analysis showed accurate boundary delineation even with cell overlap and background clutter. Conclusions: These findings indicate that the proposed framework provides a promising and scalable framework for WBC segmentation. While the current study focuses on segmentation, future work will investigate integration with classification and radiomics pipelines, as well as validation on more diverse clinical datasets, including bone marrow and leukemia samples. Full article

Background and Objectives: Lung ultrasound (LUS) has fundamentally transformed neonatal respiratory diagnostics, offering a radiation-free, bedside-applicable modality capable of guiding surfactant therapy, characterizing pulmonary pathology, and monitoring treatment response in real time. While surfactant replacement therapy is firmly established for neonatal respiratory distress syndrome (RDS), its role in acute complications—specifically pulmonary hemorrhage (PH) and pneumothorax (PTX)—remains uncertain and heterogeneous in clinical practice. This review examines how LUS-based phenotyping can improve the diagnostic precision and therapeutic sequencing of surfactant administration in these high-risk scenarios, and how comorbidities such as hemodynamically significant patent ductus arteriosus, persistent pulmonary hypertension, sepsis, and coagulopathy modulate clinical outcomes. Materials and Methods: We conducted a structured narrative review of studies published from 2020 onward, sourced from PubMed, Web of Science, Semantic Scholar, and Mendeley, using PRISMA-inspired selection principles. The search combined terms including “lung ultrasound,” “neonatal POCUS,” “surfactant therapy,” “pulmonary hemorrhage,” “neonatal pneumothorax,” and “LUS score.” Studies focusing on neonatal populations, clinical LUS applications, and surfactant use in PH and PTX were prioritized. Results: Quantitative LUS scoring systems (range 0–18) predict surfactant need and re-dosing with AUC values of 0.85–0.87, outperforming clinical estimates alone. In PH, LUS reveals dense consolidation with alveolar flooding patterns, guiding the timing of rescue surfactant after hemodynamic stabilization; response monitoring via serial LUS is feasible and informative. In PTX, hallmark signs—absent lung sliding, loss of B-lines, and the pathognomonic lung point—allow diagnosis within seconds, guiding immediate thoracentesis and subsequent surfactant administration if underlying RDS is confirmed. Nationally implemented LUS protocols in neonatal intensive care units have demonstrated significant reductions in radiation exposure without compromising diagnostic accuracy. Conclusions: LUS-guided decision algorithms—integrating ultrasonographic phenotyping, quantitative scoring, and hemodynamic assessment—represent the current best framework for individualizing surfactant therapy in neonatal PH and PTX. Standardization of POCUS training and protocol implementation in neonatal units is essential. Prospective multicenter trials are urgently needed to define optimal indications, timing, and dosing in these vulnerable populations. Full article

Convolutional neural networks (CNNs) and training-free CNN variants have been successfully applied to hyperspectral image (HSI) processing and analysis. Training-free CNNs have shown promising feature extraction performance, which could effectively address the issue of typical CNNs being highly parameterized; however, inevitable noise and redundancy in the randomly selected training-free convolutional kernels often leads to unsatisfactory performance. To address this issue, we propose Superpixel Random Selection Random Walk Multi-Branch Depthwise Convolutional Neural Network (SRSRWMD-CNN). Specifically, we propose a novel training-free convolutional neural network characterized by inter-layer multi-scale integration and intra-layer grouping. Various superpixels groups are first generated through multi-scale superpixel segmentation algorithms, then the predetermined number of superpixels are randomly sampled from these groups to serve as training-free convolution kernels. This mechanism enables adaptive computation of HSI feature maps without costly model training in the feature extraction stage, allowing the network to effectively capture a multi-scale spectral–spatial feature representation. Additionally, we propose a multi-branch depthwise convolution strategy that mitigates feature learning errors while significantly enhancing feature representation capabilities. A random walk strategy is employed to expand the receptive field and enhance the robustness of the training-free convolution kernels. Finally, the multi-scale spectral–spatial features are concatenated with the multiple convolutional stages to fuse salient shallow and deep features for accurate HSI classification. Extensive experiments demonstrate that the proposed method achieves superior performance compared to state-of-the-art algorithms. Full article

The rapid expansion of Unmanned Aircraft Systems (UAS) operations has created an urgent need for scalable strategic conflict resolution methods within the U-space framework. When requested 4D flight plans overlap with previously authorised ones, the Flight Authorisation Service denies the request and can provide the UAS operator with an alternative, conflict-free route. While traditional pathfinding algorithms ensure optimal routes, their computational cost creates a critical bottleneck during the flight activation phase or emergency missions, which demand near-instantaneous responses. To address this, we propose a three-stage framework. First, an Octree spatial partitioning discretises the airspace to identify occupied cells. Second, both A* and JPS algorithms are implemented to establish an optimal reference route. Finally, a standard Deep Reinforcement Learning (DRL) model, trained on realistic PX4 Simulator trajectories and using a well-adjusted reward function, generates alternative paths that optimise distance and energy. Results demonstrate that this DRL architecture achieves near-optimal routing behaviour. Crucially, it reduces computation time by several orders of magnitude compared to traditional algorithms, solving complex conflicts in milliseconds rather than seconds. We conclude that simple, well-tuned DRL architectures overcome latency limitations of classical pathfinding while achieving optimal results, ensuring rapid, safe, and efficient conflict resolution for high-density U-space. Full article

Background: Intermittent fasting (IF) is a popular dietary strategy for improving weight and cardiometabolic health. However, its effectiveness and potential risks across different adult age trajectories remain unclear. This systematic review and meta-analysis evaluated the age-specific effects of IF on body composition and cardiometabolic markers. Methods: Following PRISMA 2020 guidelines, PubMed, Scopus, and Web of Science were searched for randomized controlled trials (RCTs) up to September 2025. Participants were stratified into three cohorts: <30 years, 30–44 years, and ≥45 years. Random-effects meta-analyses and leave-one-out sensitivity analyses were conducted on body composition, lipid profiles, glycemic markers, and blood pressure. Additionally, a conservative methodological sensitivity analysis (imputed correlation r = 0.5) and subgroup analyses by fasting modality (TRF vs. intermittent energy restriction) were performed. Risk of bias was assessed using the RoB 2 tool. Results: Analysis of 28 RCTs (N = 1833) demonstrated that IF significantly reduced body weight and BMI across all age groups. Notably, subgroup analyses revealed comparable physiological responses between TRF and intermittent energy restriction modalities. Cardiometabolic adaptations were highly age-dependent. Young adults exhibited significant reductions in fasting insulin and HOMA-IR, alongside a robust reduction in fat mass. However, a significant loss of fat-free mass (FFM) was observed in both young and older cohorts. While middle-aged and older adults experienced the most pronounced improvements in triglycerides, systolic blood pressure, and insulin sensitivity, our conservative sensitivity analysis unmasked a significant elevation in low-density lipoprotein cholesterol (LDL-C) in this group, mirroring the robust LDL-C increase observed in young adults. Early middle-aged adults exhibited highly variable responses with no significant overall improvements in cardiometabolic parameters. Conclusions: IF is an effective weight-management tool, but elicits distinct, age-specific metabolic trajectories. While middle-aged and older adults derive pronounced cardiometabolic benefits, they face critical risks of lean mass depletion, necessitating a combined “IF+” strategy (adequate protein and resistance training). Crucially, the age-specific risk of LDL-C elevation dictates a mandate for vigilant lipid monitoring. Given that the certainty of evidence was rated as low to very low per GRADE criteria, these age-specific patterns should be interpreted as hypothesis-generating, warranting validation in future large-scale trials. Full article