Search Results (173)

Potato cultivation in the Ecuadorian Andes is largely manual and relies on intensive agrochemical inputs. We introduce a reproducible workflow that couples life cycle assessment (LCA) with a neural-network surrogate to enable rapid multi-impact screening of two potato management scenarios in Ecuador: (i) conventional manual management and (ii) Unmanned aerial vehicle (UAV)-based field monitoring to identify hotspots for targeted ground-based input application. Multi-category impacts are computed in OpenLCA using the environmental footprint method (EF 3.0) per kilogram of potatoes and scaled to annual national totals using reported national production data. UAV operation is parameterized as 0.51 kg CO₂ eq·h⁻¹, equivalent to 0.225 kg CO₂ eq·ha⁻¹ at a coverage rate of 2.27 ha·h⁻¹. For 2024, the UAV-informed scenario reduces climate change from 4.29 × 10⁷ to 3.75 × 10⁷ kg CO₂ eq (−12.7%), resource use, fossils from 5.09 × 10⁸ to 4.54 × 10⁸ MJ (−10.7%), and freshwater eutrophication from 3.33 × 10⁴ to 2.83 × 10⁴ kg P eq (−15.0%), while land use remains nearly unchanged at ~4.73 × 10⁹ Pt (−0.1%). To avoid repeated LCA recalculations, a multi-output artificial neural network (ANN) surrogate (29 outputs) was trained in Python (TensorFlow/Keras) and evaluated using leave-one-year-out (LOYO) cross-validation (2015–2024), showing strong agreement with the LCA results. This framework enables scalable what-if analysis and efficient evaluation of UAV-enabled precision monitoring strategies in resource-constrained settings. Full article

Background: Corneal organoids derived from pluripotent stem cells have emerged as powerful tools for studying corneal development, disease modeling, and regenerative medicine applications. While previous protocols have successfully generated corneal tissue structures, there remains a need for three-dimensional models that recapitulate the complex cellular architecture and diversity of native human cornea. Methods: We developed a modified spontaneous three-dimensional corneal organoid model using human embryonic stem cells (hESCs) through an adapted Self-formed Ectoderm Autonomous Multi-zone (SEAM) protocol. hESCs were cultured as spheroids in ultra-low-binding plates under normoxic conditions and differentiated over 7–8 weeks. Organoids were characterized using immunofluorescence staining for corneal-specific markers and single-cell RNA sequencing to assess cellular composition and gene expression patterns. Results: Approximately 20% of organoids developed transparent regions characteristic of corneal tissue by day 30 of differentiation. Immunofluorescence analysis revealed spatially organized expression of corneal markers, including ZO-1 and E-cadherin in the outermost epithelial layers, P63α-positive putative limbal stem cells at the epithelial–stromal interface, vimentin-positive stromal cells in the interior, and laminin-1 deposition that suggests Bowman’s membrane formation. The organoids expressed cornea-specific keratins (K3, K12, and K15) and the master regulator PAX6 in appropriate cellular compartments. Single-cell RNA sequencing identified 18 distinct cell clusters, including three corneal epithelium subclusters with differential expression of MUC16, KRT12, and ΔNp63α, two stromal populations with distinct inflammatory profiles, and a corneal endothelium cluster. Transcriptomic analysis confirmed expression of key corneal genes, including AQP3, CDH1, multiple keratins, mucins, and extracellular matrix components (HAS2, CD34, CD44, COL8A1, and KERA). Conclusions: This three-dimensional spheroid-based putative corneal organoid model successfully recapitulates the multilayered architecture and cellular diversity of human cornea, including stratified epithelium, putative limbal stem cells, stroma, and endothelium in spatially appropriate arrangements. The model demonstrates molecular signatures consistent with native corneal tissue and provides a valuable platform for studying corneal development, disease mechanisms, and potential therapeutic applications. Future optimization to improve organoid formation efficiency and functional maturation will enhance the utility of this system for both basic research and translational medicine. Full article

Deep learning has become central to semantic segmentation of three-dimensional medical images. However—despite many published models—their adoption in practice remains limited, as deployment often requires advanced programming skills and familiarity with specific machine learning frameworks. Thus, technical barriers restrict its use to specialized users. To address this, we present SegMed (version 1.0), an open-source, standalone desktop application that provides an end-to-end workflow for deep learning-based medical image segmentation. SegMed supports the loading and inspection of common medical image formats, as well as array-based formats. The application integrates standard preprocessing operations often used in the field and directly supports loading of pretrained segmentation models implemented in both PyTorch (version 2.X) and Keras (version 2.X) and those created using the Medical Open Network for AI framework (version 1.X). Models are automatically inspected to infer required configurations, such as input size and post-processing steps, enabling segmentation with minimal user intervention. Results can be exported as volumetric images or 3D surface meshes for downstream analysis, visualization, or special applications such as virtual reality. SegMed was tested using multiple publicly available pretrained models, demonstrating robustness and flexibility across diverse segmentation tasks. By abstracting low-level implementation details, SegMed lowers technical barriers, promotes reproducibility, and facilitates the integration of AI-assisted segmentation into medical imaging workflows. Full article

This study evaluates the application of deep learning techniques to the design of a microwave transmitter–receiver system operating in the 5G mid-band. The proposed architecture consists of four stages—signal generation, amplification, mixing, and filtering—each initially designed using conventional microwave methods and subsequently integrated into a complete transceiver. Simulation data were generated and component-specific convolutional neural networks (CNNs) were implemented in Python using TensorFlow/Keras. Across all models, an average error reduction exceeding 90% was achieved, with most networks converging after the third training cycle. System-level integration shows that the baseline design achieved a transmitted power of −32.637 dBm and a gain of 1.116 dB, while the deep learning-based design yielded −33.912 dBm and 0.738 dB. Additional analysis of S-parameters confirms acceptable impedance matching and a frequency response of around 3.5 GHz. These results illustrate that deep learning provides an effective complementary methodology for multi-component microwave system modeling and optimization in 5G applications. Full article

Weed identification and quantification are processes that are usually manual, subjective, and error-prone. Weeds compete with crops for nutrients, minerals, physical space, sunlight, and water. Thus, weed identification is a crucial component of precision agriculture for autonomous removal and site-specific treatments, efficient weed control, and sustainability. Convolutional Neural Networks (CNNs) are very common in weed identification. This work implemented CNN models for semantic segmentation based on the U-Net architecture for automatically segmenting and quantifying weeds in potato crops using RGB images acquired by a drone at 9–10 m height, flying at 1 m/s. Remote sensing images are affected by factors that degrade image quality and the model’s accuracy. Five U-Net variants were evaluated: the original U-Net, Residual U-Net, Double U-Net, Modified U-Net, and AU-Net. The models were trained using the TensorFlow/Keras frameworks on Google Colab Pro+, following the Knowledge Discovery in Databases (KDD) methodology for image analysis. Each model was trained using a diverse custom dataset in uncontrolled environments, considering six classes: background, Broadleaf dock (Rumex obtusifolius), Dandelion (Taraxacum officinale), Kikuyu grass (Cenchrus clandestinum), other weed species, and the crop potato (Solanum tuberosum L.). The models’ segmentation was widely assessed using Mean Dice Coefficient, Mean IoU, and Dice Loss metrics. The results showed that the Residual U-Net model performed the best in multi-class segmentation, achieving a Mean IoU of 0.8021, a performance comparable to or superior to that reported by other authors. Additionally, a Student’s t-test was applied to complement the data analysis, suggesting that the model is reliable for weed quantification. Full article

Real-time trajectory forecasting for curling stones is essential for on-ice decision support, yet prior work often emphasizes offline analysis, fixed-window predictors, or physics-driven models that require additional measurements, and it rarely reports end-to-end feasibility under edge-computing constraints (latency and memory). This leaves a practical gap between accurate trajectory reconstruction and deployable rink-side guidance. To bridge this gap, we propose an online forecaster based on low-dimensional $(x,y)$ coordinate streams and a lightweight attention-enhanced Long Short-Term Memory (LSTM) architecture optimized for edge devices. The model uses a four-second sliding window (240 frames at 59.94 Hz) to predict fifteen seconds of future positions (900 frames) in a single multi-step forward pass, and an overlapping publication scheme is adopted to retain longer temporal context and stabilize continuous updates. We further provide a TensorFlow Lite (TFLite) conversion and quantization workflow to support on-device inference. Quantitatively, experiments on the CurlTracer dataset (1033 throws at 59.94 Hz) show that the proposed attention–LSTM achieves trajectory-level MAE/MdAE of 0.25/0.22 m over the full prediction horizon, improving over a plain LSTM (0.30/0.24 m) and a physics-based pivot-slide baseline (3.52/3.54 m). At two checkpoints, the first-step MAE/MdAE are 0.14/0.11 m and the mid-step MAE/MdAE are 0.21/0.18 m. For real-time feasibility, on a Raspberry Pi 4B the per-window latency is approximately 0.25 s (including I/O and post-processing), while CPU benchmarks show that TFLite variants provide 7–8× speedups over the original Keras runtime with only minor accuracy loss (e.g., window-level MAE 0.30–0.41 m across FP32/DRQ/FP16/INT8). Qualitatively, representative trajectory visualizations show good agreement in near/mid horizons and reasonable stopping-region guidance, supporting integration with a stone-mounted interface for actionable feedback. Full article

By integrating deep learning techniques with a multi-model voting mechanism, this study aimed to predict the risk of student suspension and dropout. Conducted at the College of Informatics, Chaoyang University of Technology in Taiwan, the research utilized the AutoKeras automated machine learning framework and student data from academic years 2019 to 2023 (academic year (AY) 108–112) for model training. A majority voting strategy was employed to aggregate predictions from multiple models. To address class imbalance within the dataset, random undersampling was applied to achieve a more balanced distribution. Features from the second semester of AY 112 were used to predict enrollment status for the first semester of AY 113. Experimental results demonstrated that models trained exclusively on AY 108–112 data outperformed those trained on a broader dataset spanning AY 100–112, with the F1-score improving from 16.67 to 19.05%. Further enhancement through the ensemble voting mechanism led to an increase in overall accuracy from 66.67 to 73%, precision from 10.53 to 12.09%, and the F1-score to 21.36%. The proposed predictive model serves as an effective early warning system for identifying students at risk of suspension or dropout, thereby enabling timely counseling interventions and contributing to improved student retention. Full article

The reliability and efficiency of power system operations, especially in smart grid scenarios, depend on accurate load demand forecasting. Electrical load forecasting is crucial for power system design, fault protection and diversification as it reduces operating costs while enhancing the system’s overall reliability, stability, and efficiency from an economic and technical perspective. Previously, load forecasting analysis has frequently been limited by inadequate feature engineering and insufficient model tuning. Prediction reliability was reduced by many previous methods’ inabilities to accurately evaluate short-term variations over time and the impact of important variables. These constraints encouraged us to develop a more reliable and thorough forecasting procedure. This research proposes an enhanced short-term load forecasting framework based on a hyperparameter-tuned long short-term memory (LSTM) using a deep learning method recurrent neural network (RNN), alongside more neural network-based models such as artificial neural networks, k-nearest neighbors, and backpropagation neural networks. Hyperparameter optimization techniques (Keras Tuner, Grid SearchCV, Scikeras + Randomized SearchCV, etc.) were used to systematically tune training parameters, learning rates, and network architectures for each forecasting model to increase model accuracy. To provide a more reliable and accurate evaluation of forecasting performance, this research employs the use of an hourly load dataset (2003–2014) enhanced with historical and environmental variables. Significant statistical metrics, such as a mean absolute error of 0.0048, root mean squared error of 0.0091, coefficient of determination of R² 0.9958, and mean absolute percentage error of 1.60%, demonstrate that the hyperparameter optimized with hourly data performed better than both conventional and other deep learning models, with the highest efficiency of all tested models. In accordance with the results, accurate LSTM-RNN parameter modification significantly improves prediction accuracy. Full article

Characterization of propeller blade vibrations is essential to ensure aerodynamic performance, minimize noise emissions, and maintain structural integrity in aerospace and unmanned aerial vehicle applications. Conventional high-fidelity finite-element and fluid–structure simulations yield precise modal predictions but incur prohibitive computational costs, limiting rapid design exploration. This paper introduces a data-driven surrogate modeling framework based on a feedforward neural network to predict natural vibration frequencies of propeller blades with high accuracy and a dramatically reduced runtime. A dataset of 1364 airfoil geometries was parameterized, meshed, and analyzed in ANSYS 2024 R2 across a range of rotational speeds and boundary conditions to generate modal responses. A TensorFlow/Keras model was trained and optimized via randomized search cross-validation over network depth, neuron counts, learning rate, batch size, and optimizer selection. The resulting surrogate achieves $R^2>0.90$ and $NRMSE<0.08$ for the second and higher-order modes, while reducing prediction time by several orders of magnitude compared to full finite-element workflows. The proposed approach seamlessly integrates with CAD/CAE pipelines and supports rapid, iterative optimization and real-time decision support in propeller design. Full article

Passive acoustic monitoring (PAM) using hydrophones enables collecting acoustic data to be collected in large and diverse quantities, necessitating the need for a reliable automated classification system. This paper presents a data-processing pipeline and a set of neural networks designed for a humpback-whale-detection system. A collection of audio segments is compiled using publicly available audio repositories and extensively curated via manual methods, undertaking thorough examination, editing and clipping to produce a dataset minimizing bias or categorization errors. An array of standard data-augmentation techniques are applied to the collected audio, diversifying and expanding the original dataset. Multiple neural networks are designed and trained using TensorFlow 2.20.0 and Keras 3.13.1 frameworks, resulting in a custom curated architecture layout based on research and iterative improvements. The pre-trained model MobileNetV2 is also included for further analysis. Model performance demonstrates a strong dependence on both feature representation and network architecture. Mel spectrogram inputs consistently outperformed MFCC (Mel-Frequency Cepstral Coefficients) features across all model types. The highest performance was achieved by the pretrained MobileNetV2 using mel spectrograms without augmentation, reaching a test accuracy of 99.01% with balanced precision and recall of 99% and a Matthews correlation coefficient of 0.98. The custom CNN with mel spectrograms also achieved strong performance, with 98.92% accuracy and a false negative rate of only 0.75%. In contrast, models trained with MFCC representations exhibited consistently lower robustness and higher false negative rates. These results highlight the comparative strengths of the evaluated feature representations and network architectures for humpback whale detection. Full article

This work presents the development and evaluation of Artificial Intelligence (AI) models for the automatic classification of brain tumors in Magnetic Resonance Imaging (MRI) scans. Several deep learning architectures were implemented and compared, including VGG-19, ResNet50, EfficientNetB3, Xception, MobileNetV2, DenseNet201, InceptionV3, Vision Transformer (ViT), and an Ensemble model. The models were developed in Python (version 3.12.4) using the Keras and TensorFlow frameworks and trained on a public Brain Tumor MRI dataset containing 7023 images. Data augmentation and hyperparameter optimization techniques were applied to improve model generalization. The results showed high classification performance, with accuracies ranging from 89.47% to 98.17%. The Vision Transformer achieved the best performance, reaching 98.17% accuracy, outperforming traditional Convolutional Neural Network (CNN) architectures. Explainable AI (XAI) methods Grad-CAM, LIME, and Occlusion Sensitivity were employed to assess model interpretability, showing that the models predominantly focused on tumor regions. The proposed approach demonstrated the effectiveness of AI-based systems in supporting early diagnosis of brain tumors, reducing analysis time and assisting healthcare professionals. Full article

Unconfined Compressive Strength (UCS) of stabilized soils is commonly used for evaluating the effectiveness of soil improvement techniques. Achieving target UCS values through conventional trial-and-error approaches requires extensive laboratory experiments, which are time-consuming and resource-intensive. Automated Machine Learning (AutoML) frameworks offer a promising alternative by enabling automated, reproducible, and accessible predictive modeling of UCS values from more readily obtainable index and physical soil and stabilizer properties, reducing the reliance on experimental testing and empirical relationships, and allowing systematic exploration of multiple models and configurations. This study evaluates the predictive performance of five state-of-the-art AutoML frameworks (i.e., AutoGluon, AutoKeras, FLAML, H2O, and TPOT) using analyses of results from 10 experimental datasets comprising 2083 samples from laboratory experiments spanning diverse soil types, stabilizers, and experimental conditions across many countries worldwide. Comparative analyses revealed that FLAML achieved the highest overall performance (average PI score of 0.7848), whereas AutoKeras exhibited lower accuracy on complex datasets; AutoGluon , H2O and TPOT also demonstrated strong predictive capabilities, with performance varying with dataset characteristics. Despite the promising potential of AutoML, prior research has shown that fully automated frameworks have limited applicability to UCS prediction, highlighting a gap in end-to-end pipeline automation. The findings provide practical guidance for selecting AutoML tools based on dataset characteristics and research objectives, and suggest avenues for future studies, including expanding the range of AutoML frameworks and integrating interpretability techniques, such as feature importance analysis, to deepen understanding of soil–stabilizer interactions. Overall, the results indicate that AutoML frameworks can effectively accelerate UCS prediction, reduce laboratory workload, and support data-driven decision-making in geotechnical engineering. Full article

Breast cancer (BC) mortality rates remain high in Low- and Middle-Income Countries (LMICs) due to limited awareness, poverty, and inadequate medical facilities that hinder early detection. Although deep learning models have achieved high accuracy in BC detection (BCD), they require substantial computational resources, making them unsuitable for deployment in remote or rural areas. This study proposes a lightweight convolutional neural network (CNN) using Knowledge Distillation (KD) for BCD, where a large Teacher Model (TM) transfers learned representations to a smaller Student Model (SM), which is better suited for deployment on low-power devices. We compare it with two prominent model compression techniques: pruning and quantization. Experimental results indicate that the TensorFlow Lite (TFLite)-optimized Student Model (SM_TFLite) achieved 97.67% accuracy, representing a 2.33% relative loss to its teacher, a result comparable to other compression techniques. Its mean accuracy is 73.97% with a 95% Confidence Interval of [65.04%, 82.90%] in a cross-dataset experiment. However, SM_TFLite was the most compact (5.21 kB) and fastest (3.3 ms latency), outperforming both pruned (2924.31 kB, 13.68 ms) and quantized models (746–751 kB, 4–5 ms). Evaluation on a Raspberry Pi 4 Model B demonstrated that all models exhibited similar CPU and memory usage, with SM_TFLite causing only a minor increase in device temperature. These results demonstrate that KD combined with TFLite conversion offers the best trade-off between accuracy, compactness, and speed. Full article

We develop predictive models for OH maser occurrence in Galactic star-forming regions by integrating dense-clump physical properties from the APEX Telescope Large Area Survey of the Galaxy (ATLASGAL) and Herschel Infrared Galactic Plane Survey (Hi-GAL) 360° catalogs with maser detections and non-detections compiled in the MaserDB.net database. We compare two predictive modeling approaches for Galactic OH maser incidence: a Generalized Linear Model (GLM; logistic regression) and a compact Keras-based binary neural network (BNN). For the 1665/1667 MHz lines, both models achieve recall of 90% with a precision of approximately 50%, while for the excited-state 6031/6035 MHz lines, precision reaches roughly 20% at the same recall. We found no statistically significant difference between the BNN and GLM in out-of-sample performance. This implies that maser occurrence may be expressed as a monotonic trend without requiring nonlinear interactions. Across different catalogs and transition lines, luminosity, luminosity-to-mass ratio ($L/M$), dust temperature, and ${\mathrm{H}}_2$ column, surface, and volume densities are the most influential features for maser prediction. These variables support a physical picture in which radiative pumping favors warm, luminous, and compact clump environments. We provide an accessible online tool that allows users to predict the likelihood of OH maser emission toward ATLASGAL or Hi-GAL sources based on coordinate lists. Full article

Forest fires pose a serious threat to the environment with the potential of causing ecological harm, financial losses, and human casualties. While research suggests that climate change will increase the frequency and severity of these fires, recent developments in deep learning and convolutional neural networks (CNN) have greatly enhanced fire detection techniques and capability. These models can be leveraged by unmanned aerial vehicles (UAVs) to automatically monitor burning areas. However, drones can carry only limited computational and power resources; therefore, on-board computing capabilities are constrained by hardware limitations. This work focuses on the design of segmentation models to identify and localize active burning areas from aerial RGB images processed on limited computing resources. To achieve this goal, the research compares the performance of different variants of the DeepLabv3 neural network model for fire segmentation when trained and tested with the FLAME dataset using a k-fold cross validation approach. Experimental results are compared with U-Net, a benchmark model used with the FLAME dataset, by implementing this model in the same codebase as the DeepLabv3 model. This work demonstrates that a refined version of DeepLabv3, with a MobileNetv2 backbone using pretrained layers and a simplified atrous spatial pyramid pooling (ASPP) module, yields a similar performance to U-Net, with a precision of 87.8% and a recall of 83.2%, while only requiring 20% of the number of parameters involved with the U-Net topology. This significantly reduces memory and power consumption, enabling longer UAV flight duration and reducing the processing overhead associated with sensor input, making it more suitable for deployment on unmanned aerial vehicles. The model’s compact architecture, implemented using TensorFlow and Keras for model design and training, along with OpenCV for image preprocessing, makes it portable and easy to integrate with edge devices such as NVIDIA Jetson boards. Full article