Browse Articles
Discover research articles across all indexed journals
High-performance scene classification in remote sensing imagery using a custom deep CNN architecture
Abstract This research presents a deep novel Convolutional Neural Network (CNN) architecture specifically designed for multi-class image categorization in remote sensing data. The proposed model is evaluated using both the NWPU-RESISC45 and UC Merced Land Use datasets, each containing 10 class categories: harbor, chaparral, tennis court, industrial area, parking lot, forest, beach, overpass, airplane, and baseball diamond. Extensive testing demonstrates that the proposed CNN architecture outperforms five popular pre-trained CNN models in terms of accuracy and efficiency. Quantitative results show that the proposed model achieves an accuracy of 0.9428 on the NWPU-RESISC45 dataset and 0.93 on the UC Merced dataset. The recall scores are 0.94 and 0.93, while precision values reach 0.95 and 0.94, respectively. Furthermore, the Intersection over Union (IoU) scores are 0.89 and 0.86, while the F1-scores are 0.94 and 0.93, confirming the robustness of the model across the datasets. In terms of computational efficiency, the model demonstrates competitive training times: 3,692 s for NWPU-RESISC45 (5,057 images across 10 classes) with GPU memory usage of 12.7 gigabytes (GB) and 559 s for the UC Merced dataset (722 images across 10 classes ). To ensure and enhance the interpretability and explainability of the model’s predictions, two interpretability techniques were incorporated: Shapley Additive Explanations (SHAP) and Class Activation Mapping (CAM). The key novelties of this manuscript include: a hybrid CNN framework that not only advances classification performance but also incorporates explainability via SHAP and CAM, while maintaining efficient training times and strong generalization, making it a compelling solution for remote sensing image understanding. In general, the key novel aspects of our approach are: Lightweight and efficient architecture: Our proposed CNN design strikes a balance between performance and computational efficiency, making it highly suitable for real-time or resource-constrained environments without compromising accuracy. Integrated interpretability: By incorporating both SHAP (Shapley Additive explanations) and CAM (Class Activation Mapping), our framework delivers strong predictive performance. Generalizability across datasets: We demonstrate the model’s robustness and generalizability across two challenging and diverse datasets, confirming its adaptability to different domain scenarios.
Study on optimization of multimodal transportation path of Jiamusi grain considering cargo loss under low carbon policy
The early impact of bariatric surgery on metabolic dysfunction-associated steatotic liver disease (MASLD) as assessed by fibroscan at 6 months postoperatively
Enhancement of yield and fruit quality of strawberry under deficit fertigation via sodium hydrosulfide and L-cysteine
Non-local attention enhanced deep learning for robust cyberattack detection in industrial IoT-based SCADA systems
Abstract Industrial Internet of Things (IIoT)-enabled Supervisory Control and Data Acquisition (SCADA) systems are pivotal for real-time monitoring and control in critical sectors like energy, manufacturing, and water management. However, their connectivity and complexity expose them to cyber threats, including zero-day vulnerabilities and advanced persistent threats (APTs). Traditional security measures, like signature-based intrusion detection systems (IDSs), are inadequate against dynamic attacks. This study introduces DeepNonLocalNN, a deep learning model combining convolutional neural networks (CNNs) with non-local attention blocks to capture local patterns and global dependencies in IIoT network traffic. Evaluated on the WUSTL-IIoT-2021 dataset, DeepNonLocalNN achieved strong performance, with an accuracy of 0.9999, a receiver operating characteristic-area under the curve (ROC-AUC) of 1.0000, and a macro F1-score of 0.93, outperforming baseline models such as NonLocalNN, CNNWithAttention, ResidualAttentionNetwork, and Long Short-Term Memory (LSTM). Notably, it excelled in detecting minority attack classes, including Backdoor (F1: 0.73) and Command Injection (CommInj, F1: 0.92), addressing class imbalance. The model’s scalable architecture, leveraging non-local attention and regularization, provides a high-performance solution for SCADA security in IIoT environments. Future work will focus on adapting the DeepNonLocalNN approach to real-time intrusion detection. It also aims to reduce the computational cost for resource-constrained PLCs and RTUs in SCADA systems. We also aim to validate this model on various industrial datasets and SCADA environments.
Sustainable fishing
Rheological investigation of the effects of tea polyphenols on SBS-modified bitumen before and after short-term aging
Impact of seaweeds on tensile, thermal and viscoelasticity behavior of polybutylene adipate terephthalate-based composites
The structural, mechanical, electrical, and radiation-shielding properties of newly yttrium and neodymium-doped lithium-zinc-phosphate glasses
Loss of mechanical stress induces synovitis, fibrosis and articular cartilage degeneration via distinct synovial cell subsets
ReFaceX: donor-driven reversible face anonymisation with detached recovery
Abstract Organisations must share facial imagery that remains useful for analysis while protecting identity. Many current methods fail to strike this balance: reconstruction-centred encoder–decoder designs tend to blur salient detail, whereas latent edits in pretrained generators often retain or drift identity cues, undermining privacy and utility. We present ReFaceX, a reversible anonymisation framework that separates what to protect from what to preserve. A donor identity code steers a U-Net anonymiser with Identity Feature Fusion to change identity while retaining non-identity content such as pose, background and expression. A learned steganographic channel carries a compact recovery payload, and reconstruction gradients are blocked at the stego image so the anonymiser is never rewarded for keeping identity. The threat model is stated explicitly and outcomes are audited with strong recognisers. On LFW and CelebA-HQ datasets at $$256\times 256$$ , ReFaceX reduces identity similarity across FaceNet, ArcFace and AdaFace, and improves recovered-image quality (SSIM $$0.9378$$ , LPIPS $$0.1002$$ , PSNR $$23.97$$ dB), while operating in real time on a single RTX 3090. Robustness to common JPEG re-encoding is also demonstrated. By turning the privacy–utility balance into an explicit and auditable operating choice, ReFaceX provides a practical template for responsible release of facial imagery and a foundation for extensions to video, higher resolutions and stronger recovery guarantees.
An image compression-encryption algorithm based on BP neural network optimized with fireworks algorithm
Surgical delays between indication and operating room access in patients undergoing glaucoma filtration surgery
Abstract Surgical delays between indication for surgery and access to the operating room in glaucoma filtration surgery (GFS) are unknown. We reviewed medical charts of the first fifty patients’ undergoing GFS from February 2017, 2019, and 2022 with the aim to: (i) measure waiting times between indication for surgery to pre-surgical workup, pre-surgical workup to surgery, and indication to surgery; (ii) identify factors affecting the pre-surgical path duration, and (iii) evaluate whether waiting times changed in the 2017–2022 quinquennium. 633 patients, in four tertiary-care Italian Centers, were enrolled. At the indication for surgery, the mean deviation (MD) was − 13.4 dB (IQR: -21.2; -6.9), with an advanced glaucoma in 54.6% of cases (MD: -20.3 dB), and a median IOP of 24 mmHg (IQR: 20.0–28.0). Overall, patients waited 44.0 days (IQR: 21.0–72.0) between indication for GFS and surgery, with the interval between indication and pre-surgical workup being the most time consuming step (32.0 days (IQR: 8.0-51.8)). Patients living in South Italy, with primary glaucoma, an IOP less than 20 mmHg, scheduled for a first phaco-combined trabeculectomy, non-monocular, and with systemic comorbidities, waited more. Since glaucoma may continue to progress while waiting, efficient organizational strategies should be adopted to optimize the pre-surgical path duration.
Structural and textural characterization of Brassica carinata biochar to investigate its potential industrial applications
Assessment of self-management empowerment program on sense of coherence, self-efficacy and postpartum anxiety in cesarean mothers randomized trial
Enhancing Arabic healthcare fake news detection with data augmentation and multi-metric analysis using large language models
Abstract The spread of fake news about healthcare can result in a global health crisis, as it is easy to mislead the public. Detection of fake Arabic news in the healthcare sector is crucial for identifying disinformation, especially in regions where Arabic is the predominant language. Various deep learning and machine learning methods have been proposed to categorize false Arabic news related to healthcare. However, the linguistic diversity of Arabic complicates the development of effective models. Furthermore, the lack of domain-specific high-quality data makes it difficult to build accurate and reliable models. Data augmentation (DA) techniques have shown great promise in addressing these challenges. This study presents a novel technique for expanding Arabic healthcare data by conducting a multi-metric analysis to comprehensively evaluate the quality of the augmented data based on several key aspects, including label preservation, novelty, diversity, and semantic similarity. In the initial phase of our research, we investigated the impact of various data augmentation techniques on widely used classification algorithms. Additionally, similarity thresholds are systematically examined to determine their effect on the classification task. Cosine and Jaccard distances are employed to evaluate the generated sentences in terms of semantics, diversity, novelty, and label preservation. Finally, we propose a novel ensemble augmentation approach that combines multiple DA techniques to generate more varied data. Based on the overall experimental results, the proposed methodology significantly improves the classification of Arabic fake news using AraBERT, with an accuracy increase of 12.1%. In comparison, Random Forest achieved an improvement of 14.7%.