Search Results (3,929)

The accurate detection of small objects remains a critical challenge in autonomous driving systems, where improving detection performance typically comes at the cost of increased model complexity, conflicting with the lightweight requirements of edge deployment. To address this dilemma, this paper proposes LEAD-YOLO (Lightweight Efficient Autonomous Driving YOLO), an enhanced network architecture based on YOLOv11n that achieves superior small object detection while maintaining computational efficiency. The proposed framework incorporates three innovative components: First, the Backbone integrates a lightweight Convolutional Gated Transformer (CGF) module, which employs normalized gating mechanisms with residual connections, and a Dilated Feature Fusion (DFF) structure that enables progressive multi-scale context modeling through dilated convolutions. These components synergistically enhance small object perception and environmental context understanding without compromising network efficiency. Second, the neck features a hierarchical feature fusion module (HFFM) that establishes guided feature aggregation paths through hierarchical structuring, facilitating collaborative modeling between local structural information and global semantics for robust multi-scale object detection in complex traffic scenarios. Third, the head implements a shared feature detection head (SFDH) structure, incorporating shared convolution modules for efficient cross-scale feature sharing and detail enhancement branches for improved texture and edge modeling. Extensive experiments validate the effectiveness of LEAD-YOLO: on the nuImages dataset, the method achieves 3.8% and 5.4% improvements in mAP@0.5 and mAP@[0.5:0.95], respectively, while reducing parameters by 24.1%. On the VisDrone2019 dataset, performance gains reach 7.9% and 6.4% for corresponding metrics. These findings demonstrate that LEAD-YOLO achieves an excellent balance between detection accuracy and model efficiency, thereby showcasing substantial potential for applications in autonomous driving. Full article

The strength mobilisation framework was adopted for the first time to describe the stress–strain responses for three different types of sands, including a total of 30 published drained triaxial tests—25 for Karlsruhe Fine Sand, 2 for Ottawa sands and 3 for Fontainebleau sand, under confining pressures ranging from 50 to 400 kPa. The peak shear strength ${\tau }_{peak}$ obtained from drained triaxial shearing of these sands was used to normalise shear stress. Shear strains normalised at peak strength ${\gamma }_{peak}$ and at half peak of shear strength ${\gamma }_{M=2}$ were taken as the normalised reference strains, and the results were compared. Power–law functions were then derived when the mobilised strength was between $0.2{\tau }_{peak}$ and $0.8{\tau }_{peak}$. Exponents of the power–law functions of these sands were found to be lower than in the published undrained shearing data of clays. Using ${\gamma }_{M=2}$ as the reference strain shows a slightly better power–law correlation than using ${\gamma }_{peak}$. Linear relationships between the reference strains and variables, such as relative density, relative dilatancy index, and dilatancy, are identified. Full article

The impact mechanism of long-term creep in gas-containing coal on coal and gas outbursts has not been fully elucidated and remains insufficiently understood for the purpose of disaster engineering control. This investigation conducted triaxial creep experiments on raw coal specimens under controlled confining pressures, axial stresses, and gas pressures. Through systematic analysis of coal’s physical responses across different loading conditions, we developed and validated a novel creep damage constitutive model for gas-saturated coal through laboratory data calibration. The key findings reveal three characteristic creep regimes: (1) a decelerating phase dominates under low stress conditions, (2) progressive transitions to combined decelerating–steady-state creep with increasing stress, and (3) triphasic decelerating–steady–accelerating behavior at critical stress levels. Comparative analysis shows that gas-free specimens exhibit lower cumulative strain than the 0.5 MPa gas-saturated counterparts, with gas presence accelerating creep progression and reducing the time to failure. Measured creep rates demonstrate stress-dependent behavior: primary creep progresses at 0.002–0.011%/min, decaying exponentially to secondary creep rates below 0.001%/min. Steady-state creep rates follow a power law relationship when subject to deviatoric stress (R² = 0.96). Through the integration of Burgers viscoelastic model with the effective stress principle for porous media, we propose an enhanced constitutive model, incorporating gas adsorption-induced dilatational stresses. This advancement provides a theoretical foundation for predicting time-dependent deformation in deep coal reservoirs and informs monitoring strategies concerning gas-bearing strata stability. This study contributes to the theoretical understanding and engineering monitoring of creep behavior in deep coal rocks. Full article

Plant disease classification is critical for effective crop management. Recent advances in deep learning, especially Vision Transformers (ViTs), have shown promise due to their strong global feature modeling capabilities. However, ViTs often overlook local features and suffer from feature extraction degradation during patch merging as channels increase. To address these issues, we propose PLTransformer, a hybrid model designed to symmetrically capture both global and local features. We design a symmetric multi-scale convolutional module that combines two different-scale receptive fields to simultaneously extract global and local features so that the model can better perceive multi-scale disease morphologies. Additionally, we propose an overlap-attentive channel downsampler that utilizes inter-channel attention mechanisms during spatial downsampling, effectively preserving local structural information and mitigating semantic loss caused by feature compression. On the PlantVillage dataset, PLTransformer achieves 99.95% accuracy, outperforming DeiT (96.33%), Twins (98.92%), and DilateFormer (98.84%). These results demonstrate its superiority in handling multi-scale disease features. Full article

CT scanners are essential tools in modern medical imaging. Sudden failures of their X-ray tubes can lead to equipment downtime, affecting healthcare services and patient diagnosis. However, existing prediction methods based on a single model struggle to adapt to the multi-stage variation characteristics of tube lifespan and have limited modeling capabilities for temporal features. To address these issues, this paper proposes an intelligent prediction architecture for CT tubes’ remaining useful life based on a dual-branch neural network. This architecture consists of two specialized branches: a residual self-attention BiLSTM (RSA-BiLSTM) and a multi-layer dilation temporal convolutional network (D-TCN). The RSA-BiLSTM branch extracts multi-scale features and also enhances the long-term dependency modeling capability for temporal data. The D-TCN branch captures multi-scale temporal features through multi-layer dilated convolutions, effectively handling non-linear changes in the degradation phase. Furthermore, a dynamic phase detector is applied to integrate the prediction results from both branches. In terms of optimization strategy, a dynamically weighted triplet mixed loss function is designed to adjust the weight ratios of different prediction tasks, effectively solving the problems of sample imbalance and uneven prediction accuracy. Experimental results using leave-one-out cross-validation (LOOCV) on six different CT tube datasets show that the proposed method achieved significant advantages over five comparison models, with an average MSE of 2.92, MAE of 0.46, and R² of 0.77. The LOOCV strategy ensures robust evaluation by testing each tube dataset independently while training on the remaining five, providing reliable generalization assessment across different CT equipment. Ablation experiments further confirmed that the collaborative design of multiple components is significant for improving the accuracy of X-ray tubes remaining life prediction. Full article

The image represents the post-mortem heart of a 28-year-old female patient, diagnosed in childhood with complete common atrioventricular canal defect. At time of diagnosis, the family refused surgery, as did the patient during her adulthood. Despite being advised against pregnancy, she became pregnant. On presentation to hospital, she was cyanotic, with clubbed fingers, and hemodynamically unstable, in sinus rhythm, with Eisenmenger syndrome and respiratory failure partially responsive to oxygen. During pregnancy, owing to systemic vasodilatation, the right-to-left shunt is increased, with more severe cyanosis and low cardiac output. Echocardiography revealed the complete common atrioventricular canal defect, with a single atrioventricular valve with severe regurgitation, right ventricular hypertrophy, pulmonary artery dilatation, severe pulmonary hypertension and a hypoplastic left ventricle. The gestational age at delivery was 38 weeks. She gave birth to a healthy boy, with an Apgar score of 10. The vaginal delivery was chosen by an interdisciplinary team. The cesarean delivery and the anesthesia were considered too risky compared to vaginal delivery. Three days later, the patient died. The autopsy revealed hepatomegaly, a greatly hypertrophied right ventricle with a purplish clot ascending the dilated pulmonary arteries and a hypoplastic left ventricle with a narrowed chamber. A single valve was observed between the atria and ventricles, making all four heart chambers communicate, also insufficiently developed interventricular septum and its congenital absence in the cranial third. These morphological changes indicate the complete common atrioventricular canal defect, with right ventricular dominance, which is a rare and impressive malformation that requires mandatory treatment in early childhood in order for the condition to be solved. Full article

Background/Objectives: Pneumatic dilation (PD) is a widely used treatment modality in the management of achalasia. It is particularly relevant in regions where many centers lack access to advanced therapeutic modalities. Therefore, we aimed to assess the effectiveness and safety of PD in our local region. Methods: This study retrospectively analyzed patients with achalasia that underwent PD from 1/2013 to 12/2019. The diagnosis of achalasia was established on the grounds of clinical symptoms, radiological and endoscopic findings, and esophageal manometry. Data on patient’s clinical characteristics, dilation technique and postprocedural follow-up were collected and statistically analyzed. Procedure effectiveness was defined as the postprocedural Eckardt score ≤ 3. Results: PD significantly reduced frequency of dysphagia, regurgitation, and retrosternal pain (p < 0.001). Body-weight increased significantly one month and one year after the procedure (p < 0.001). The procedural success rate was 100%. No severe complications were reported. Conclusions: PD is an effective and safe treatment modality in the management of achalasia. The study limitations include a single center design with the small number of participants, not all of whom underwent manometry, gender disproportion, absence of non-responders, and a short follow-up. Full article

This article presents VRF-DETR, a lightweight real-time object detection framework for aerial remote sensing images, aimed at addressing the challenge of insufficient receptive fields for easily confused categories due to differences in height and perspective. Based on the RT-DETR architecture, our approach introduces three key innovations: the multi-scale receptive field adaptive fusion (MSRF²) module replaces the Transformer encoder with parallel dilated convolutions and spatial-channel attention to adjust receptive fields for confusing objects dynamically; the gated multi-scale context (GMSC) block reconstructs the backbone using Gated Multi-Scale Context units with attention-gated convolution (AGConv), reducing parameters while enhancing multi-scale feature extraction; and the context-guided fusion (CGF) module optimizes feature fusion via context-guided weighting to resolve multi-scale semantic conflicts. Evaluations were conducted on both the VisDrone2019 and UAVDT datasets, where VRF-DETR achieved the mAP₅₀ of 52.1% and the mAP_50-95 of 32.2% on the VisDrone2019 validation set, surpassing RT-DETR by 4.9% and 3.5%, respectively, while reducing parameters by 32% and FLOPs by 22%. It maintains real-time performance (62.1 FPS) and generalizes effectively, outperforming state-of-the-art methods in accuracy-efficiency trade-offs for aerial object detection. Full article

Background: Transcatheter aortic valve implantation (TAVI) has revolutionized the treatment of severe aortic valve stenosis (AS). Balloon post-dilatation (PD) remains an important procedural step to optimize valve function by resolving incomplete valve expansion, which may lead to paravalvular regurgitation and other potentially adverse effects. There are only limited data on the predictors, incidence, and clinical impact of PD during TAVI. Methods: This retrospective, single-center study analyzed 585 patients who underwent TAVI (2016–2022). Pre-procedural evaluations included transthoracic echocardiography and CT angiography to assess key parameters, including the aortic valve calcium score (AVCS); aortic valve calcium density (AVCd); aortic valve maximal systolic transvalvular flow velocity (AV Vmax); and aortic valve mean systolic pressure gradient (AV MPG). We identified imaging predictors of PD and evaluated associated clinical outcomes by analyzing procedural endpoints (according to VARC-3 criteria) and long-term survival. Results: PD was performed on 67 out of 585 patients, with elevated AV Vmax (OR: 1.424, 95% CI: 1.039–1.950; p = 0.028) and AVCd (OR: 1.618, 95% CI: 1.227–2.132; p = 0.001) emerging as a significant independent predictor for PD in TAVI. Kaplan–Meier survival analysis revealed no significant differences in short- and mid-term survival between patients who underwent PD and those who did not. Interestingly, patients requiring PD exhibited a lower incidence of adverse events regarding major vascular complications, permanent pacemaker implantations and stroke. Conclusions: The study highlights AV Vmax and AVCd as key predictors of PD. Importantly, PD was not associated with increased procedural adverse events and did not predict adverse events in this contemporary cohort. Full article

Because artificially cemented granular (ACG) materials employ diverse combinations of aggregates and binders—including cemented soil, low-cement-content cemented sand and gravel (LCSG), and concrete—their stress–strain responses vary widely. In LCSG, the binder dosage is typically limited to 40–80 kg/m³ and the sand–gravel skeleton is often obtained directly from on-site or nearby excavation spoil, endowing the material with a markedly lower embodied carbon footprint and strong alignment with current low-carbon, green-construction objectives. Yet, such heterogeneity makes a single material-specific constitutive model inadequate for predicting the mechanical behavior of other ACG variants, thereby constraining broader applications in dam construction and foundation reinforcement. This study systematically summarizes and analyzes the stress–strain and volumetric strain–axial strain characteristics of ACG materials under conventional triaxial conditions. Generalized hyperbolic and parabolic equations are employed to describe these two families of curves, and closed-form expressions are proposed for key mechanical indices—peak strength, elastic modulus, and shear dilation behavior. Building on generalized plasticity theory, we derive the plastic flow direction vector, loading direction vector, and plastic modulus, and develop a concise, transferable elastoplastic model suitable for the full spectrum of ACG materials. Validation against triaxial data for rock-fill materials, LCSG, and cemented coal–gangue backfill shows that the model reproduces the stress and deformation paths of each material class with high accuracy. Quantitative evaluation of the peak values indicates that the proposed constitutive model predicts peak deviatoric stress with an error of 1.36% and peak volumetric strain with an error of 3.78%. The corresponding coefficients of determination R² between the predicted and measured values are 0.997 for peak stress and 0.987 for peak volumetric strain, demonstrating the excellent engineering accuracy of the proposed model. The results provide a unified theoretical basis for deploying ACG—particularly its low-cement, locally sourced variants—in low-carbon dam construction, foundation rehabilitation, and other sustainable civil engineering projects. Full article

Background: Right ventricular (RV) dysfunction is associated with poor clinical outcomes in critically ill sepsis patients, but its pathophysiology and predictors are incompletely characterized. We aimed to investigate the predictors of RV dysfunction and its outcomes in sepsis patients admitted to the intensive care unit (ICU). Methods: This is a single-center retrospective cohort study of adult patients admitted to the ICU for sepsis who had echocardiography within 72 h of diagnosis. Patients with acute coronary syndrome, acute decompensated heart failure, or significant valvular dysfunction were excluded. RV dysfunction was defined as the presence of RV dilation, hypokinesis, or both. Demographics and clinical outcomes were obtained from electronic medical records. Results: A total of 361 patients were included in our study—47 with and 314 without RV dysfunction. The mean age of the population was 66.8 years and 54.6% were females. Compared to those without RV dysfunction, patients with RV dysfunction were more likely to require mechanical ventilation (63.8% vs. 43.9%, p = 0.01) and vasopressor support (61.7% vs. 36.6%, p < 0.01). On multivariate logistic regression analysis, increasing age (OR 1.03, 95% C.I. 1.00–1.06), a history of HIV infection (OR 5.88, 95% C.I. 1.57–22.11) and atrial fibrillation (OR 4.34, 95% C.I. 1.83–10.29), and presence of LV systolic dysfunction (OR 14.40, 95% C.I. 5.63–36.84) were independently associated with RV dysfunction. Patients with RV dysfunction had significantly worse 30-day survival (Log-Rank p = 0.023). On multivariate Cox regression analysis, older age (HR 1.02, 95% C.I. 1.00–1.04) and peak lactate (HR 1.16, 95% C.I. 1.11–1.21) were independent predictors of 30-day mortality. Conclusions: Among other findings, our data suggests a possible association between a history of HIV infection and RV dysfunction in critically ill sepsis patients, and this should be investigated further in future studies. Patients with evidence of RV dysfunction had poorer survival in this population; however this was not an independent predictor of mortality in the multivariate analysis. A larger cohort with a longer follow-up period may provide further insights. Full article

Accurate cloud segmentation is critical for astronomical observations and solar forecasting. However, traditional threshold- and texture-based methods suffer from limited accuracy (65–80%) under complex conditions such as thin cirrus or twilight transitions. Although the deep-learning segmentation method based on U-Net effectively captures low-level and high-level features and achieves significant progress in accuracy, current methods still lack interpretability and multi-scale feature integration and usually produce fuzzy boundaries or fragmented predictions. In this paper, we propose multi-scale CAM, an explainable AI (XAI) framework that integrates class activation mapping (CAM) with hierarchical feature fusion to quantify pixel-level attention across hierarchical features, thereby enhancing the model’s discriminative capability. To achieve precise segmentation, we integrate CAM into an improved U-Net architecture, incorporating multi-scale CAM attention for adaptive feature fusion and dilated residual modules for large-scale context extraction. Experimental results on the SWINSEG dataset demonstrate that our method outperforms existing state-of-the-art methods, improving recall by 3.06%, F1 score by 1.49%, and MIoU by 2.21% over the best baseline. The proposed framework balances accuracy, interpretability, and computational efficiency, offering a trustworthy solution for cloud detection systems in operational settings. Full article

The main objective of this study was to compare the long-term outcomes of transcatheter aortic valve implantation (TAVI) in patients with severe aortic stenosis, focusing on differences between self-expanding valve (SEV) versus balloon-expandable valve (BEV) prostheses and the influence of balloon pre- and post-dilatation on clinical results. The secondary objective was to report the long-term outcomes after TAVI in Romania. All patients who underwent a TAVI procedure for severe AS between November 2016 and May 2025 at a tertiary center in Romania were included in the present study. A total of 702 patients were included, of which 455 (64.8%) and 247 (35.1%) patients received a BEV (Sapien3 platform) and a SEV (Accurate, Boston, Portico, Evolut, or Navitor platforms), respectively. Pre-dilatation was performed in 514 (73.2%) cases, and post-dilatation was performed in 189 (26.9%) cases. There were 10.5 and 7.8 all-cause and cardiovascular-cause mortality event rates per 100 patient years, respectively. In regard to the univariable Cox regression, a BEV has significantly lower mortality than an SEV (HR = 0.67[0.46–0.96], p = 0.03), pre-dilatation did not influence mortality (HR = 0.71[0.48–1.04], p = 0.08), and post-dilatation significantly increased mortality (HR = 1.51[1.05–2.19], p = 0.03). In regard to the multivariable Cox regression, survival was not influenced by pre-dilatation or the valve platform, while post-dilatation had a trend towards higher mortality (p = 0.06). The BEV and SEV have similar survival rates, with no heterogeneity among a large number of TAVI platforms. While pre-dilatation had no impact on mortality, post-dilatation was associated with a trend towards increased mortality (p = 0.06), which was independent of the transprosthetic gradient. Survival after TAVI in Romania is comparable to that reported in Western registries. Full article

In this research project a large linear electromagnetic actuator (LLEA) was designed and manufactured. The electromagnetic performance was published in previous works, but in this paper we focus on the technological challenges related to the manufacturing in particular. This LLEA was based on the magnet-free switched-reluctance principle, having six effective energised stator “teeth” and four passive mover parts (4:6 ratio). Various aspects and challenges encountered during the manufacturing, transport, and assembly are discussed. Thermal expansion of steel contributed to the decision of the modular design, with each module having 1.3 m in length, with a 2 mm longitudinal dilatation gap. The initial prototype was tested with a 10.6 m length, with plans to extend the test track to 60 m, which was fully achievable due to the modular design and required 29 tons of electrical steel to be built. The stator laminations were cut by a bespoke progressive tool with stamping, and other parts by a CO₂ laser. Mounting was based on welding (back of the stator) and clamping plates (through insulated bolts). The linear longitudinal force was on the order of 8 kN, with the main air gap of 7.5–10 mm on either side of the mover. The lateral forces could exceed 40 kN and were supported by appropriate construction steel members bolted to the concrete floor. The overall mechanical tolerances after installation remained below 0.5 mm. The technology used for constructing this prototype demonstrated the cost-effective way for a semi-industrial manufacturing scale. Full article

In the era of smart agriculture, accurate identification of plant species is critical for effective crop management, biodiversity monitoring, and the sustainable use of medicinal resources. However, existing deep learning approaches often underperform when applied to fine-grained plant classification tasks due to the lack of domain-specific, high-quality datasets and the limited representational capacity of traditional architectures. This study addresses these challenges by introducing a novel, well-curated leaf image dataset consisting of 39 classes of medicinal and aromatic plants collected from the Aromatic and Medicinal Plant Research Station in Odakkali, Kerala, India. To overcome performance bottlenecks observed with a baseline Convolutional Neural Network (CNN) that achieved only 44.94% accuracy, we progressively enhanced model performance through a series of architectural innovations. These included the use of a pre-trained VGG16 network, data augmentation techniques, and fine-tuning of deeper convolutional layers, followed by the integration of Squeeze-and-Excitation (SE) attention blocks. Ultimately, we propose a hybrid deep learning architecture that combines VGG16 with Batch Normalization, Gated Recurrent Units (GRUs), Transformer modules, and Dilated Convolutions. This final model achieved a peak validation accuracy of 95.24%, significantly outperforming several baseline models, such as custom CNN (44.94%), VGG-19 (59.49%), VGG-16 before augmentation (71.52%), Xception (85.44%), Inception v3 (87.97%), VGG-16 after data augumentation (89.24%), VGG-16 after fine-tuning (90.51%), MobileNetV2 (93.67), and VGG16 with SE block (94.94%). These results demonstrate superior capability in capturing both local textures and global morphological features. The proposed solution not only advances the state of the art in plant classification but also contributes a valuable dataset to the research community. Its real-world applicability spans field-based plant identification, biodiversity conservation, and precision agriculture, offering a scalable tool for automated plant recognition in complex ecological and agricultural environments. Full article