Search Results (3,316)

To combat the decline in substation image clarity resulting from adverse weather phenomena like haze, which often leads to poor illumination and altered color perception, a compact image dehazing model called the Substation Image Enhancement Network with Decomposition Convolution and Adaptive Fusion (SDCNet) is introduced. In contrast to traditional dehazing methods that expand the convolutional kernel to widen the receptive field and improve feature acquisition, commonly at the cost of increased parameters and computational load, SDCNet employs a decomposition-based convolutional enhancement module. This component efficiently extracts spatial features while keeping computation lightweight. Moreover, an adaptive fusion mechanism is incorporated to better align and merge features from both encoder and decoder stages, aiding in the retention of essential image information. To further enhance model learning, a contrastive regularization strategy is applied, leveraging both hazy and clear substation images during training. Empirical evaluations show that SDCNet substantially enhances visual brightness and restores accurate structural and color details. On the MIIS dataset of substation haze images, it delivers gains of 4.053 dB in PSNR and 0.006 in SSIM compared to current state-of-the-art approaches. Additional assessment on the SSDF dataset further confirms its reliability in detecting substation defects under unfavorable weather conditions. Full article

Most existing deep learning-based super-resolution (SR) methods for remote sensing images rely on predefined degradation assumptions (e.g., bicubic downsampling). However, when real-world degradations deviate from these assumptions, their performance deteriorates significantly. Moreover, explicit degradation estimation approaches based on iterative schemes inevitably lead to accumulated estimation errors and time-consuming processes. In this paper, instead of explicitly estimating degradation types, we first innovatively introduce an MSCN_G coefficient to capture global prior information corresponding to different distortions. Subsequently, distortion-enhanced representations are implicitly estimated through contrastive learning and embedded into a super-resolution network equipped with multiple distortion decoders (D-Decoder). Furthermore, we propose a distortion-related channel segmentation (DCS) strategy that reduces the network’s parameters and computation (FLOPs). We refer to this Global Prior-guided Distortion-enhanced Representation Learning Network as GDRNet. Experiments on both synthetic and real-world remote sensing images demonstrate that our GDRNet outperforms state-of-the-art blind SR methods for remote sensing images in terms of overall performance. Under the experimental condition of anisotropic Gaussian blurring without added noise, with a kernel width of 1.2 and an upscaling factor of 4, the super-resolution reconstruction of remote sensing images on the NWPU-RESISC45 dataset achieves a PSNR of 28.98 dB and SSIM of 0.7656. Full article

Emotion recognition in speech is essential for enhancing human–computer interaction (HCI) systems. Despite progress in Bangla speech emotion recognition, challenges remain, including low accuracy, speaker dependency, and poor generalization across emotional expressions. Previous approaches often rely on traditional machine learning or basic deep learning models, struggling with robustness and accuracy in noisy or varied data. In this study, we propose a novel multi-stream deep learning feature fusion approach for Bangla speech emotion recognition, addressing the limitations of existing methods. Our approach begins with various data augmentation techniques applied to the training dataset, enhancing the model’s robustness and generalization. We then extract a comprehensive set of handcrafted features, including Zero-Crossing Rate (ZCR), chromagram, spectral centroid, spectral roll-off, spectral contrast, spectral flatness, Mel-Frequency Cepstral Coefficients (MFCCs), Root Mean Square (RMS) energy, and Mel-spectrogram. Although these features are used as 1D numerical vectors, some of them are computed from time–frequency representations (e.g., chromagram, Mel-spectrogram) that can themselves be depicted as images, which is conceptually close to imaging-based analysis. These features capture key characteristics of the speech signal, providing valuable insights into the emotional content. Sequentially, we utilize a multi-stream deep learning architecture to automatically learn complex, hierarchical representations of the speech signal. This architecture consists of three distinct streams: the first stream uses 1D convolutional neural networks (1D CNNs), the second integrates 1D CNN with Long Short-Term Memory (LSTM), and the third combines 1D CNNs with bidirectional LSTM (Bi-LSTM). These models capture intricate emotional nuances that handcrafted features alone may not fully represent. For each of these models, we generate predicted scores and then employ ensemble learning with a soft voting technique to produce the final prediction. This fusion of handcrafted features, deep learning-derived features, and ensemble voting enhances the accuracy and robustness of emotion identification across multiple datasets. Our method demonstrates the effectiveness of combining various learning models to improve emotion recognition in Bangla speech, providing a more comprehensive solution compared with existing methods. We utilize three primary datasets—SUBESCO, BanglaSER, and a merged version of both—as well as two external datasets, RAVDESS and EMODB, to assess the performance of our models. Our method achieves impressive results with accuracies of 92.90%, 85.20%, 90.63%, 67.71%, and 69.25% for the SUBESCO, BanglaSER, merged SUBESCO and BanglaSER, RAVDESS, and EMODB datasets, respectively. These results demonstrate the effectiveness of combining handcrafted features with deep learning-based features through ensemble learning for robust emotion recognition in Bangla speech. Full article

With the increasing demand for sustainable and cost-effective alternatives in the construction industry, polymer composites have emerged as a promising solution. This study focuses on the development of innovative composite tiles using Linear Low-Density Polyethylene (LLDPE) powder blended with epoxy resin and a hardener as a green substitute for conventional ceramic and cement tiles. LLDPE is recognized for its flexibility, durability, and chemical resistance, making it an effective filler within the epoxy matrix. To optimize its material properties, composite samples were fabricated using three different LLDPE-to-epoxy ratios: 30:70, 40:60, and 50:50. Flexural strength testing revealed that while the 50:50 blend achieved the highest maximum value (29.887 MPa), it also exhibited significant variability, reducing its reliability for practical applications. In contrast, the 40:60 ratio demonstrated more consistent and repeatable flexural strength, ranging from 16 to 20 MPa, which is ideal for flooring applications where mechanical performance under repeated loading is critical. Scanning Electron Microscopy (SEM) images confirmed uniform filler dispersion in the 40:60 mix, further supporting its mechanical consistency. The 30:70 composition showed irregular and erratic behaviour, with values ranging from 11.596 to 25.765 MPa, indicating poor dispersion and increased brittleness. To complement the development of the materials, deep learning techniques were employed for real-time defect detection in the manufactured tiles. Utilizing the YOLOv8 (You Only Look Once version 8) algorithm, this study implemented an automated, vision-based surface monitoring system capable of identifying surface deterioration and defects. A dataset comprising over 100 annotated images was prepared, featuring various surface defects such as cracks, craters, glaze detachment, and tile lacunae, alongside defect-free samples. The integration of machine learning not only enhances quality control in the production process but also offers a scalable solution for defect detection in large-scale manufacturing environments. This research demonstrates a dual approach to material innovation and intelligent defect detection to improve the performance and quality assurance of composite tiles, contributing to sustainable construction practices. Full article

The increasingly severe issue of marine debris presents a critical threat to the sustainable development of marine ecosystems. Real-time detection is essential for timely intervention and cleanup. Furthermore, the density of marine debris exhibits significant depth-dependent variation, resulting in degraded detection accuracy. Based on 9625 publicly available underwater images spanning various depths, this study proposes UTNet, a lightweight neural model, to improve the effectiveness of real-time intelligent identification of marine debris through multidimensional optimization. Compared to Faster R-CNN, SSD, and YOLOv5/v8/v11/v12, the UTNet model demonstrates enhanced performance in random image detection, achieving maximum improvements of 3.5% in mAP50 and 9.3% in mAP50-95, while maintaining reduced parameter count and low computational complexity. The UTNet model is further evaluated on underwater videos for real-time debris recognition at varying depths to validate its capability. Results show that the UTNet model exhibits a consistently increasing trend in confidence levels across different depths as detection distance decreases, with peak values of 0.901 at the surface and 0.764 at deep-sea levels. In contrast, the other six models display greater performance fluctuations and fail to maintain detection stability, particularly at intermediate and deep depths, with evident false positives and missed detections. In summary, the lightweight UTNet model developed in this study achieves high detection accuracy and computational efficiency, enabling real-time, high-precision detection of marine debris at varying depths and ultimately benefiting mitigation and cleanup efforts. Full article

This paper investigates the suppression of the cohesive cracking mode (CCM) in the sintered silver (s-Ag) die layer by intentionally introducing anisotropic porosity through two press sintering methods. Full press (FP) and local press (LP) bonding represent the s-Ag formed by pressing the die-attached assemblies (DAAs) on either the entire top surface or only on the silicon carbide (SiC) top surface, respectively. The fabricated DAAs were encapsulated with epoxy molding compounds. Degradation was evaluated using a nine-point bending test (NBT) under cyclic force between 0 and 270 N with a triangle waveform for 3 min per cycle at 150 °C. Scanning tomography images after 500 NBT cycles showed that the LP reduced the inner degradation ratio by up to 21.1% compared to the FP. Cross-sectional scanning electron microscopy revealed that the FP progressed cracking in the s-Ag die layer, whereas the LP showed no evidence of cracking. A finite element analysis revealed that in the FP, the accumulated plastic strain (APS) was concentrated in the s-Ag layer within the inner SiC chip. In contrast, the APS of the LP was preferentially concentrated outside the SiC chip. This preferential localization of damage outside the chip presents a promising approach for enhancing the reliability of packaging products. Full article

Background/Objectives: This study aimed to evaluate the diagnostic performance of simplified intravoxel incoherent motion (IVIM) diffusion-weighted imaging (DWI) parameters in distinguishing malignant from benign breast lesions, and to explore their association with clinicopathological features. Methods: This retrospective study included 108 women who underwent breast MRI with multi-b-value DWI (0, 20, 200, 500, 800 s/mm²). Of those 108 women, 73 had pathologically confirmed malignant lesions. IVIM maps (ADC_map, D, D*, and perfusion fraction f) were generated using IB-Diffusion™ software version 21.12. Lesions were manually segmented by radiologists, and clinicopathological data including receptor status, Ki-67 index, cancer type, histologic grade, and molecular subtype were extracted from medical records. Nonparametric tests and ROC analysis were used to assess group differences and diagnostic performance. Additionally, a binary logistic regression model combining D, D*, and f was developed to evaluate their joint diagnostic utility, with ROC analysis applied to the model’s predicted probabilities. Results: Malignant lesions demonstrated significantly lower diffusion parameters compared to benign lesions, including ADC_map (p = 0.004), D (p = 0.009), and D* (p = 0.016), indicating restricted diffusion in cancerous tissue. In contrast, the perfusion fraction (f) did not show a significant difference (p = 0.202). ROC analysis revealed moderate diagnostic accuracy for ADC_map (AUC = 0.671), D (AUC = 0.657), and D* (AUC = 0.644), while f showed poor discrimination (AUC = 0.576, p = 0.186). A combined logistic regression model using D, D*, and f significantly improved diagnostic performance, achieving an AUC of 0.725 (p < 0.001), with 67.1% sensitivity and 74.3% specificity. ADC_map achieved the highest sensitivity (100%) but had low specificity (11.4%). Among clinicopathological features, only histologic grade was significantly associated with IVIM metrics, with higher-grade tumors showing lower ADC_map and D* values (p = 0.042 and p = 0.046, respectively). No significant associations were found between IVIM parameters and ER, PR, HER2 status, Ki-67 index, cancer type, or molecular subtype. Conclusions: Simplified IVIM DWI offers moderate accuracy in distinguishing malignant from benign breast lesions, with diffusion-related parameters (ADC_map, D, D*) showing the strongest diagnostic value. Incorporating D, D*, and f into a combined model enhanced diagnostic performance compared to individual IVIM metrics, supporting the potential of multivariate IVIM analysis in breast lesion characterization. Tumor grade was the only clinicopathological feature consistently associated with diffusion metrics, suggesting that IVIM may reflect underlying tumor differentiation but has limited utility for molecular subtype classification. Full article

An 89-year-old male developed a persistent high fever (around 39 °C) approximately two weeks following endoscopic reduction of sigmoid volvulus. He had no history of hypercalcemia but was using diuretics and proton pump inhibitors. Renal and thyroid status were normal. He was largely bedridden and asymptomatic except for fever. Laboratory tests demonstrated elevated C-reactive protein (4.75 mg/dL), but some tumor markers (including CEA, CA19-9, and CA125), anti-nuclear antibodies, MPO-ANCA, PR3-ANCA, β-D-glucan, and interferon-gamma release assay were all negative. Urinalysis was unremarkable. Blood cultures obtained from two sets were negative. Chest–abdomen–pelvis contrast-enhanced computed tomography (CT), and echocardiography did not reveal any evident neoplastic lesions or focal sites of infection. Despite various antibiotic therapies, the patient’s spike fever persisted for nearly one month, leading to a diagnosis of fever of unknown origin (FUO). The patient experienced partial symptomatic relief with corticosteroid therapy, though mild fever continued. Two months after the volvulus onset, diffusion-weighted whole-body imaging with background body signal suppression (DWIBS) was performed, revealing hyperintensities at the right sternoclavicular joint, leading to a diagnosis of sternoclavicular arthritis. Neck CT revealed calcification in this joint. Despite difficulty in joint fluid analysis, low infection risk and the patient’s prolonged bedridden state and advanced age led to suspicion of pseudogout. Nonsteroidal anti-inflammatory drugs relieved fever and normalized inflammatory markers. DWIBS may be a valuable tool for detecting potential focus sites in FUO. Full article

Regulating tumor invasion and metastasis is pivotal for improving cancer patient prognosis. While cell migration is a key factor in these processes, the non-targeted effects of chemoradiotherapy on cell motility remain poorly understood. In this study, we employed HeLa-FUCCI cells—a cervical cancer-derived HeLa cell line integrated with the Fluorescent Ubiquitination-Based Cell Cycle Indicator (FUCCI) probe, enabling the visualization of cell cycle phases—to investigate the radiation-induced impacts, including non-targeted effects, on cell migration. To create irradiated (In-field) and non-irradiated (out-of-field) regions, half of the culture dish was shielded with a lead block during irradiation. Cells were then exposed to 2 Gy X-rays, with or without cisplatin. Following irradiation, the cells were subjected to time-lapse imaging at 15 min intervals for 24 h, and the acquired data were analyzed using cell segmentation and tracking algorithms, Cellpose 2.0 and TrackMate 7. Without cisplatin, the migration velocity and total distance traveled of Out-of-field cells were significantly reduced compared to controls, suggesting a suppressive bystander signal. In contrast, with cisplatin treatment, these parameters significantly increased in both In-field and Out-of-field cells. This suggests that chemoradiotherapy may inadvertently enhance tumor cell motility outside the target volume, a critical finding with significant implications for therapeutic outcomes. Full article

The visual cryptography scheme (VCS) is a fundamental image encryption technique that divides a secret image into two or more shares, such that the original image can be revealed by superimposing a sufficient number of shares. A major limitation of conventional VCS methods is pixel expansion, wherein the generated shares and reconstructed image are typically at least twice the size of the original. Additionally, thin lines or curves—only one pixel wide in the original image—often appear distorted or duplicated in the reconstructed version, a distortion known as the thin-line problem (TLP). To eliminate the reliance on predefined codebooks inherent in traditional VCS, Kafri introduced the Random Grid visual cryptography scheme (RG-VCS), which eliminates the need for such codebooks. This paper introduces novel algorithms that are among the first to explicitly address the thin-line problem in the context of random grid based schemes. This paper presents novel visual cryptography algorithms specifically designed to address the thin-line preservation problem (TLP), which existing methods typically overlook. A comprehensive visual and numerical comparison was conducted against existing algorithms that do not explicitly handle the TLP. The proposed methods introduce adaptive encoding strategies that preserve fine image details, fully resolving TLP-2 and TLP-3 and partially addressing TLP-1. Experimental results show an average improvement of 18% in SSIM and 13% in contrast over existing approaches. Statistical t-tests confirm the significance of these enhancements, demonstrating the effectiveness and superiority of the proposed algorithms. Full article

Unmanned aerial vehicles (UAVs) play an ever-increasing role in disaster response and remote sensing. However, the deep learning models they rely on remain highly vulnerable to adversarial attacks. This paper presents an evaluation and defense framework aimed at enhancing adversarial robustness in aerial disaster image classification using the AIDERV2 dataset. Our methodology is structured into the following four phases: (I) baseline training with clean data using ResNet-50, (II) vulnerability assessment under Projected Gradient Descent (PGD) attacks, (III) adversarial training with PGD to improve model resilience, and (IV) comprehensive post-defense evaluation under identical attack scenarios. The baseline model achieves 93.25% accuracy on clean data but drops to as low as 21.00% under strong adversarial perturbations. In contrast, the adversarially trained model maintains over 75.00% accuracy across all PGD configurations, reducing the attack success rate by more than 60%. We introduce metrics, such as Clean Accuracy, Adversarial Accuracy, Accuracy Drop, and Attack Success Rate, to evaluate defense performance. Our results show the practical importance of adversarial training for safety-critical UAV applications and provide a reference point for future research. This work contributes to making deep learning systems on aerial platforms more secure, robust, and reliable in mission-critical environments. Full article

Over recent decades, the hyperspectral image (HSI) classification landscape has undergone significant transformations driven by advances in deep learning (DL). Despite substantial progress, few-shot scenarios remain a significant challenge, primarily due to the high cost of manual annotation and the unreliability of visual interpretation. Traditional DL models require massive datasets to learn sophisticated feature representations, hindering their full potential in data-scarce contexts. To tackle this issue, a deep transfer contrastive learning network is proposed. A spectral data augmentation module is incorporated to expand limited sample pairs. Subsequently, a spatial–spectral feature extraction module is designed to fuse the learned feature information. The weights of the spatial feature extraction network are initialized with knowledge transferred from source-domain pretraining, while the spectral residual network acquires rich spectral information. Furthermore, contrastive learning is integrated to enhance discriminative representation learning from scarce samples, effectively mitigating obstacles arising from the high inter-class similarity and large intra-class variance inherent in HSIs. Experiments on four public HSI datasets demonstrate that our method achieves competitive performance against state-of-the-art approaches. Full article

Quantitative dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is emerging as a valuable tool for assessing tumor and parenchymal perfusion in the liver, playing a developing role in locoregional therapies (LRTs) for hepatocellular carcinoma (HCC). This review explores the conceptual underpinnings and early investigational stages of DCE-MRI for LRTs, including thermal ablation, transarterial chemoembolization (TACE), and transarterial radioembolization (TARE). Preclinical and early-phase studies suggest that DCE-MRI may offer valuable insights into HCC tumor microvasculature, treatment response, and therapy planning. In thermal ablation therapies, DCE-MRI provides a quantitative measurement of tumor microvasculature and perfusion, which can guide more effective energy delivery and estimation of ablation margins. For TACE, DCE-MRI parameters are proving their potential to describe treatment efficacy and predict recurrence, especially when combined with adjuvant therapies. In ⁹⁰Y TARE, DCE-MRI shows promise for refining dosimetry planning by mapping tumor blood flow to improve microsphere distribution. However, despite these promising applications, there remains a profound gap between early investigational studies and clinical translation. Current quantitative DCE-MRI research is largely confined to phantom models and initial feasibility assessments, with robust retrospective data notably lacking and prospective clinical trials yet to be initiated. With continued development, DCE-MRI has the potential to personalize LRT treatment approaches and serve as an important tool to enhance patient outcomes for HCC. Full article

An enhancement method integrating dynamic color correction with a lightweight residual network is proposed to resolve the challenges of color bias and insufficient contrast in underwater imaging. The dynamic color correction module is implemented based on the gray-world assumption, adaptively adjusting inter-channel color shifts to mitigate blue-green dominance in acquired images. Subsequently, the corrected images are processed through an improved SRResNet architecture incorporating lightweight residual blocks with embedded channel–spatial attention mechanisms, enhancing the responses of feature channels and the saliency of spatial regions Model complexity is reduced through depthwise separable convolutions and channel dimension reduction, ensuring computational efficiency. Validation on UIEB and RUIE datasets demonstrates superior qualitative and quantitative performance, achieving PSNR gains of 0.92–5.95 dB and UCIQE improvements of 0.14–0.74, compared with the established methodologies. Ablation studies quantify the contributions of the color correction and attention mechanisms to the overall enhancement efficiency, verifying the network’s effectiveness. Full article

Accurate segmentation of fine roots in field rhizotron imagery is essential for high-throughput root system analysis but remains challenging due to limitations of traditional methods. Traditional methods for root quantification (e.g., soil coring, manual counting) are labor-intensive, subjective, and low-throughput. These limitations are exacerbated in in situ rhizotron imaging, where variable field conditions introduce noise and complex soil backgrounds. To address these challenges, this study develops an advanced deep learning framework for automated segmentation. We propose an improved U-shaped Convolutional Neural Network (U-Net) architecture optimized for segmenting larch (Larix olgensis) fine roots under heterogeneous field conditions, integrating both in situ rhizotron imagery and open-source multi-species minirhizotron datasets. Our approach integrates (1) a Convolutional Block Attention Module (CBAM) to enhance feature representation for fine-root detection; (2) an additive feature fusion strategy (UpAdd) during decoding to preserve morphological details, particularly in low-contrast regions; and (3) a transfer learning protocol to enable robust cross-species generalization. Our model achieves state-of-the-art performance with a mean intersection over union (mIoU) of 70.18%, mean Recall of 86.72%, and mean Precision of 75.89%—significantly outperforming PSPNet, SegNet, and DeepLabV3+ by 13.61%, 13.96%, and 13.27% in mIoU, respectively. Transfer learning further elevates root-specific metrics, yielding absolute gains of +0.47% IoU, +0.59% Precision, and +0.35% F1-score. The improved U-Net segmentation demonstrated strong agreement with the manual method for quantifying fine-root length, particularly for third-order roots, though optimization of lower-order root identification is required to enhance overall accuracy. This work provides a scalable approach for advancing automated root phenotyping and belowground ecological research. Full article