Search Results (736)

The cutting head is the primary cutting load device of the roadheader. The harsh, complex excavation environment often leads to a scarcity of labelled fault samples, significantly hindering the development of accurate fault diagnosis models. This paper proposes a novel fault diagnosis method based on the Gramian angular difference field (GADF) and an attention-enhanced transfer-learning AlexNet. The collected one-dimensional vibration signals are transformed into two-dimensional image data using GADF to capture transient impact gradients and preserve absolute temporal correlations. To overcome data limitations, the method retains the base convolutional feature extractor of an AlexNet model pre-trained on ImageNet, whilst discarding the original fully connected and classification layers. A novel classification head incorporating a multihead self-attention (MSA) mechanism is constructed to fine-tune the network specifically for the fault diagnosis task. This structural modification adaptively assigns higher weights to fault-sensitive spatial regions, significantly enhancing the model’s feature extraction focus and generalisation capability even under intense background noise. Experimental validation was conducted on a scaled cutting head fault diagnosis test bench. The results demonstrate that the proposed method outperforms other baselines across evaluation metrics, exhibiting robust recognition accuracy and stability. This effectively identifies the cutting head’s operating condition, offering a novel and practical approach for future underground fault diagnosis in coal mines. Full article

Digital computing-in-memory (DCIM) provides deterministic floating-point computation but incurs substantial area and power overhead from replicated mantissa multipliers and adder trees. This work proposes an error-recoverable BF16 DCIM arithmetic unit that jointly approximates a 2-bit multiplier and the first adder stage. For the 11 × 11 input, the multiplier outputs 0111 instead of the Baseline 1111, converting the error from +6 to −2 and fixing the product MSB to 0. This enables the first adder stage to be reduced from 4 bits to 3 bits. A lightweight flag detects the same error condition and is reused as a carry input for local compensation, avoiding a separate multi-bit correction circuit. Hierarchical design-space exploration selected the 0111 approximation with carry compensation at bit position 1. Transistor-level evaluation showed reductions of 14.81% in transistor count and 28.56% in average power relative to the Baseline. Across ResNet18, VGG16-BN, and AlexNet on CIFAR-10 and CIFAR-100, the Proposed scheme achieved the lowest BF16-referenced Layer NRMSE and Logit NRMSE among the evaluated Baseline, DIMC-S-derived, LSAC OR+SXAFA-derived, and Proposed schemes, while the Top-1 accuracy difference relative to the Baseline remained within −0.02%p to +0.12%p. These results demonstrate an improved hardware–accuracy trade-off without retraining or data rearrangement. Full article

Poly(methyl methacrylate) (PMMA) is one of the most widely used polymeric biomaterials in prosthodontics and digital dentistry because of their clinical reliability and compatibility with computer-aided design/computer-aided manufacturing (CAD/CAM). Its widespread use raises questions regarding material consumption, manufacturing waste, recyclability, and circularity. A critical narrative review supported by a structured literature search was conducted. PubMed, Scilit, OpenAlex, and ScienceDirect were searched for English-language literature published from January 2000 to June 2026. Targeted Google Scholar searches, cross-referencing, standards, and official technical sources supplemented the search. Evidence was organized according to its directness to dental PMMA and synthesized thematically. Prepolymerized CAD/CAM PMMA provides consistent material quality and generally improved mechanical performance compared with conventionally processed PMMA; however, subtractive manufacturing generates disc remnants, milling particles, and polishing residues. Mechanical recycling and depolymerization demonstrate technical recovery potential, although evidence specific to heterogeneous dental waste streams, environmental performance, and clinical-grade reuse remains limited. Technical recyclability should not be automatically equated with a viable circular economy or a net environmental benefit. Future research should quantify dental PMMA waste, establish effective collection and recovery pathways, and integrate life-cycle assessments with clinical performance and safety standards. Full article

Hyperparameter optimization is a critical determinant of the performance and generalizability of convolutional neural networks (CNNs), yet selecting optimal configurations remains a complex challenge due to the large, non-linear search space. While numerous optimization techniques have been explored in the literature, most comparative studies have relied on benchmark datasets, limiting insights into their practical effectiveness under real-world constraints. This study evaluates four widely used hyperparameter optimization methods (i.e., Grid Search, Optuna, Hyperopt, and DEAP) across five CNN architectures (i.e., AlexNet, GoogLeNet, LeNet, MobileNet, and ResNet18) under two distinct data regimes: the Modified National Institute of Standards and Technology (MNIST) benchmark and a custom SkinCancer image dataset. On MNIST, all configurations achieved strong classification performance, with test accuracies ranging from 95.45% to 97.61%, led by ResNet18 optimized via Grid Search (97.61%). In this simpler setting, DEAP provided the shortest optimization runtimes, while Optuna and Hyperopt offered an effective balance between performance and computational cost. In contrast, the SkinCancer dataset exhibited substantially greater variability in performance, highlighting the stronger influence of both model architecture and optimization strategy. ResNet18 achieved the highest overall test accuracy (81.50%) when optimized with Hyperopt. Although no single optimization method consistently outperformed all others, adaptive approaches frequently achieved competitive or superior predictive performance while requiring considerably less computation than exhaustive search. These findings demonstrate that hyperparameter optimization dynamics depend heavily on dataset complexity, where computational efficiency is the primary differentiator for simpler classification tasks, but optimization architecture selection becomes critical for navigating challenging medical imaging applications. Full article

Chinese bayberry (Myrica rubra) is an extensively popular fruit with unique flavors and high nutritional values. A precise and efficient method for assessing external quality traits of Chinese bayberry is critical for machine automatic sorting to ensure the quality and increase economic value. Here, we developed an automatic evaluation flow based on deep learning detection and a multi-trait scoring framework for postharvest quality grading of Chinese bayberry. Firstly, the YOLOv11 framework was applied to detect the Chinese bayberry fruits, and the network consists of a CSPDarknet-based backbone with C3k2 modules, an SPPF block, a bidirectional PANet neck, and three decoupled detection heads. The experimental results showed a precision of 0.9991 and a recall of 1.0000 for the waxberry class (mAP@0.5 = 0.9950, mAP@0.5:0.95 = 0.8632). Furthermore, the quality scoring model was built on a modified AlexNet backbone that provided more stable training and better accuracy compared to the other seven CNN models. The R², MAE, and RMSE showed the close alignment between predicted and reference values, revealing reliable estimations of our six fruit-quality traits, including radius, color depth, color class, color uniformity, surface injury area ratio, and plumpness. This self-developed flow provides a comprehensive and effective assessment method to improve fruit quality and economic value, which further facilitates automated agricultural production. Full article

Near-infrared (NIR) spectroscopy-based adulteration detection approaches are still dominated by one-dimensional (1D) spectral analysis, which inherently limits the exploration of complex patterns and nonlinear interactions in spectral data. Therefore, the objective of this study is to use a two-dimensional (2D) NIR spectrogram, combined with Explainable Artificial Intelligence (XAI), to predict the level of adulteration in palmyrah sap. The dataset matrix dimension is 110 × 1101, derived from the sample adulteration level (0–100%) and the NIR wavenumber (4000–12,500 cm⁻¹). Following Kennard–Stone partitioning, the evaluated preprocessing methods were applied using parameters derived exclusively from the training set. For the 2D modeling branch, the resulting training and testing spectra were subsequently transformed separately using the Continuous Wavelet Transform (CWT). A total of six AI algorithms, three from machine learning (PLS, kNN, ANN) and three from deep learning (CNN, AlexNet, ResNet), were applied in this study. The best model AI was interpreted using Shapley Additive Explanations (SHAP) for 1D NIRs and the Gradient-weighted Class Activation Mapping (Grad-CAM) for 2D NIR spectrograms. The four best-performing model configurations can predict the level of palmyrah sap adulteration, with R² values ranging from 0.969 to 0.994 and RMSE ranging from 2.333% to 5.547% in the training. In the testing, the model’s performance is in the R² range of 0.959–0.990, RMSE of 3.093–6.396%, MAE of 2.358–4.252%, RPD of 5.06–10.46 and Bias of 0.03–0.93%. The SHAP and Grad-CAM XAI revealed that the wavenumber associated with this sap counterfeiting is critical to the level of adulteration of palmyrah sap. This approach provides a quantitative method that accounts for advanced dimensions and treats them as essential information to support large-scale data matrices in AI modeling. The application of this method is an alternative that is easy to interpret and implement, and can be applied to long- and short-wavelength data from continuous NIR or discrete multi-wavelength NIR. Full article

As silicon scaling nears its physical limits and digital systems process ever-growing amounts of data, integrating computation directly within memory is emerging as a key strategy to overcome the constraints of conventional Von Neumann architectures. Approximate In-Memory Computation (IMC) with memristors offers a promising path toward energy-efficient processing for data-intensive applications. Recent adaptive approximate adders exploit operand magnitude to dynamically switch between exact and approximate computation, but typically ignore carry propagation across approximation boundaries, which can significantly degrade application-level robustness. This work introduces RapproX, a family of adaptive memristive approximate adders featuring a lightweight carry lookback mechanism that approximates carry interaction between exact and approximate regions. The proposed approach improves arithmetic robustness while introducing only minimal overhead and enabling resource-efficient implementations through memristor reuse. Experimental results demonstrate that the proposed approaches achieve superior arithmetic quality compared to State-of-the-Art (SoA) memristive approximate adders. More importantly, the carry lookback mechanism translates into substantial application-level benefits. In image processing, RapproX reduces energy consumption by up to 30.9% compared to the most competitive SoA design and by 50.3% compared to exact computation while maintaining roughly 43 dB Peak Signal-to-Noise Ratio (PSNR). Across a range of machine-learning workloads, including k-means, AlexNet on MNIST, and multiple CIFAR-10 models, RapproX preserves near-exact inference accuracy for the evaluated models at low-to-moderate k and maintains the energy advantages of adaptive approximation, while SoA approximations degrade markedly under the same conditions. These simulation-based results suggest that lightweight carry-aware approximation can improve the robustness of adaptive approximate in-memory computing with only marginal hardware overhead. Full article

Traffic accidents are one of the most significant societal problems worldwide, resulting in loss of life and property. In recent years, powerful solutions for the automatic classification of traffic accidents have been offered by deep learning-based image processing methods. This study comparatively evaluated four convolutional neural network architectures (SqueezeNet, ResNet-18, ResNet-50 and AlexNet) with different depths and levels of complexity using the CADD (Car Accidents and Deformation) dataset, which consists of images of vehicle accidents. The model’s performance was examined in detail using accuracy, precision, recall, the F1 score, a confusion matrix and an ROC–AUC analysis. The ResNet-50 model significantly outperformed all others, achieving 52% validation accuracy and balanced F1-score values (~0.50–0.52). It also achieved particularly high precision (0.70), recall (0.79) and AUC (0.922) values in the Totaled class. In contrast, SqueezeNet exhibited significant class bias and struggled to learn the multi-class structure. The AlexNet and ResNet-18 models showed moderate performance, achieving limited success in terms of discriminability, particularly in the ‘Severe’ class. It should be emphasized that the analysis relies on single-frame static images and a relatively small, predefined dataset; the study is therefore intended as a preliminary comparative benchmark rather than a deployable damage-assessment system. Within this scope, the findings indicate that residual architectures such as ResNet-50 are promising candidates for future deep learning-based accident analysis tools, although further validation on larger and more diverse datasets is required prior to their real-world use in traffic safety, insurance assessment, and intelligent transportation applications. Full article

Accurate classification of infrasound signals is essential for nuclear-test verification, natural-hazard warning, and geophysical monitoring. Conventional convolutional neural networks (CNN) applied to this task tend to overfit small, class-imbalanced datasets and cannot quantify predictive uncertainty. To address these limitations, we introduce a Bayesian CNN framework that treats network weights as probability distributions and performs inference by variational approximation. LeNet-5, AlexNet, and 4Conv3Fc network serve as baselines and are converted into Bayes LeNet-5, Bayes AlexNet, and Bayes 4Conv3Fc. The short-time Fourier transform (STFT) provides time–frequency spectrograms as model input. On a highly imbalanced dataset comprising nuclear tests, chemical explosions, volcanic eruptions, rocket launches, earthquakes, and lightning, Bayes 4Conv3Fc reaches an overall accuracy of 99.14% without any data augmentation. Relative to the deterministic baselines, precision, recall, and F1-score increase by up to 6.91, 7.12, and 7.20 percentage points, respectively, and Cohen’s Kappa coefficient by up to 8.86 percentage points. Against class-weighted cross-entropy, a standard imbalance-handling baseline, the Bayesian models yield consistently lower Brier scores, indicating that the gains stem from principled uncertainty modelling rather than loss re-weighting alone.. This study quantifies both epistemic and aleatoric uncertainty in an infrasound signal classification model, and the calibration analysis validates that these uncertainty estimates are reliable, providing a basis for evaluating prediction reliability and diagnosing potential failure modes, thereby contributing to improved model interpretability. Coupled with an event-level data partitioning strategy, the evaluation faithfully reflects the model’s generalization to unseen events and offers a promising direction toward uncertainty-aware infrasound monitoring. Full article

This tutorial takes the reader on a historical and technical journey from the simple Perceptron (1958) to modern Convolutional Neural Networks (CNNs) that dominate spatial data processing (images and video). We start with the Perceptron’s linear classifier, then expose its inability to learn non-linear patterns (e.g., XOR), which motivates the Multi-Layer Perceptron (MLP) and the backpropagation algorithm. Next, we discuss the limitations of MLP when faced with structured data like images—parameter explosion, loss of spatial information, and lack of translation invariance—and use these limitations as a natural springboard to the core ideas of CNNs: local connectivity, weight sharing, and hierarchical feature learning. Throughout, we provide intuitive explanations, mathematical formulations, and step-by-step numerical examples (e.g., a complete forward and backward pass for a small network, and a manual 2D convolution). Clear graphical representations and examples help readers understand each concept. The tutorial concludes with a detailed walkthrough of influential CNN architectures (LeNet-5, AlexNet, VGG, GoogLeNet, ResNet, DenseNet, and EfficientNet) and also discusses more recent attention-based models (e.g., Vision Transformers and ConvNeXt), explaining why each was necessary and how it advanced the field. Aimed at students and practitioners with a basic knowledge of calculus and linear algebra, this tutorial connects foundational ideas to state-of-the-art deep learning, focusing on spatial data. It is designed for readers who want to understand why each architectural choice was made, not just what the final model looks like. Full article

This study aimed to develop and evaluate a deep learning diagnostic model integrating a convolutional autoencoder (CAE) and ResNet18 for the early and accurate diagnosis of developmental dysplasia of the hip (DDH) in children, addressing the subjectivity of traditional methods. This study began with the construction of a CAE network. By introducing a multi-layer compression structure and a Dropout layer, the network was forced to learn the low-dimensional and robust feature representations of the input X-ray images (frog-leg lateral view). The CAE was used to reconstruct the input images and generate enhanced images that highlight abnormal regions. Subsequently, the original and enhanced images were concatenated along the channel dimension to form an information-rich enhanced input. Finally, the pretrained ResNet18 was adopted as the backbone classification network, and its input layer was modified to adapt to multi-channel input to conduct training and classification for the stitched images. Compared to those achieved by the four benchmark models (ResNet18, DarkNet19, AlexNet, and MobileNetV2), the CAE-ResNet18 model achieves excellent performance on the test set. The accuracy, recall rate, and F1-score of the Normal class are 0.9890, 1.0000, and 0.9945, respectively. The accuracy, recall rate, and F1-score of the DDH class were 1.0000, 0.9912, and 0.9956, respectively. Visual analysis shows that the t-SNE visualization of the fully connected layer feature of this model presents a more obvious inter-class separation. The CAE-ResNet18 model effectively leverages both original image information and abnormal features, significantly improving the accuracy and reliability of DDH diagnosis. It provides a potential intelligent tool for clinical auxiliary diagnosis, which may enhance patient prognosis. Full article

The transition toward Industry 4.0 requires the integration of technologically viable hardware–software–intelligence solutions into existing industrial infrastructures to enable smart and autonomous production systems. Thus, Industry 4.0 enables techno-symmetry, which refers to the balanced and interactive distribution of technological capacity, information processing ability, and decision-making capability across production networks. This study proposes a comprehensive hardware–software–intelligence framework for a real-time visual quality inspection of transformer cases during the manufacturing process by using an embedded deep learning architecture. First, a real-world dataset consisting of 232 defective and 264 non-defective printed transformer case images was collected from the production line of a transformer manufacturing facility and preprocessed to improve data quality and model generalization. Second, to enhance feature extraction capability, the classical AlexNet architecture was modified to develop a Multiscale AlexNet (MS-AN) model capable of simultaneously capturing both global and local spatial features. The proposed architecture incorporates parallel convolutional branches with 3 × 3 and 5 × 5 receptive fields, which are fused at the feature level to increase representation diversity and improve robustness against noise and degradation in printed images. Third, an experimental system was implemented using practical industrial automation technologies (e.g., CUDA-accelerated C++ programming, the NVIDIA Jetson Orin Nano edge computing platform, ROS-based communication infrastructure, IoT protocols, and programmable logic controller (PLC) integration). Experimental results demonstrate that the proposed system achieves real-time inspection performance of approximately 2 s per inspection with 99% classification accuracy on the constructed dataset. The developed framework enables efficient deployment of deep learning models on GPU-based edge devices; thus, it reduces reliance on workstation-class computers, lowers energy consumption, and supports scalable intelligent inspection architectures aligned with Industry 4.0 transformation objectives. Full article

It is important to detect the damage in the rollers and repair them since the damage to the rollers has a negative impact on the quality of the rolled products. Identifying the types of damage helps determine the repair process and normal production work. Ultrasonic testing technology has the advantages of large detection depth, accurate defect localization, low cost, convenient use, fast speed, and harmlessness to the human body. In order to improve the intelligence of ultrasonic detection for identifying damages in rollers, this article proposes a deep learning classification method of damages based on Rayleigh wave signals and power spectrum images with specific sampling rate, automatic identification of four common types of damages (void, hole, crack, and adhesion) is achieved by establishing end-to-end learning models for one-dimensional (1D) and two-dimensional (2D) data. Firstly, an organic glass inclined block and a clamping device were designed. In the experiment, time-domain signals were received on the right side of the damaged sample, and signal data sets were established for signals with different sampling rates. Then, the power spectrum image data sets were established after a short-time Fourier transform was performed. Next, a damage detection model is established based on a deep learning framework, which includes ResNet, GoogLeNet, DenseNet, and AlexNet with 1D and 2D convolutional channels to extract signal features for classifying damage. Finally, the performances of DenseNet models with different structures and depths are compared based on key indicators such as accuracy and training time. The experiment demonstrates that under high sampling rate conditions, using the power spectrum image of Rayleigh waves as data input yields better results than directly using Rayleigh wave signals. Moreover, for the power spectrum images of 0.5 MS/s Rayleigh waves, using ResNet-18 to establish a deep learning model can achieve high accuracy and shorter training time. Full article

This study examines the effectiveness of Genetic Algorithm (GA)-based image augmentation in enhancing Convolutional Neural Network (CNN) performance for Romblon marble texture classification. Five CNN architectures, AlexNet, InceptionV3, VGG16, MobileNet, and ResNet50, were trained and evaluated on both raw and GA-augmented datasets. Unprocessed dataset achieved limited accuracy, with InceptionV3 performing best at 35.33%, followed closely by VGG16 at 33.78%. In contrast, GA-augmented data significantly boosted performance, with VGG16 achieving 94.68% accuracy, followed by MobileNet (92.68%) and InceptionV3 (92.46%). Entropy loss values consistently decreased across all models, indicating improved convergence and reduced overfitting. Although ROC-AUC scores remained close to 0.5, reflecting modest improvements in class separability, overall results confirm that evolutionary augmentation enriches dataset diversity and strengthens CNN learning capacity. MobileNet showed a solid balance between accuracy and computational economy, underscoring the possibility of GA-based augmentation as a workable option for real-world marble categorization, while VGG16 emerged as the most accurate of the studied architectures. Full article

Historical buildings are important cultural assets that reflect the identity of cities and preserve the collective memory of societies. However, these structures are increasingly exposed to environmental degradation and human-induced impacts, making their systematic documentation and condition assessment essential for effective conservation strategies. Recent advances in artificial intelligence have provided powerful tools for image-based analysis in the field of heritage preservation. In particular, transfer learning enables the adaptation of pre-trained deep learning models to domain-specific tasks with limited labeled data. In this study, a deep transfer learning-based framework is proposed for automatic damage detection and classification in historical buildings. A new near-balanced dataset of 20,000 images spanning six deterioration categories was developed and made publicly available. Ten convolutional neural network and transformer architectures pre-trained on ImageNet were systematically compared under a unified Bayesian optimization protocol. Experimental results on a held-out test set show that EfficientNetB3 achieves the highest classification accuracy (97.65%), while AlexNet obtains the lowest performance (83.89%); the validation set was used exclusively for hyperparameter tuning. The results demonstrate that transfer learning-based models can effectively identify visually observable deterioration patterns and provide reliable support for automated documentation processes. The proposed framework contributes to the development of data-driven decision-support tools for digital documentation and condition assessment in heritage conservation. Full article