Recent Advances in Computer Science and Communications - Volume 18, Issue 8, 2025

Deep learning models, such as deep convolutional neural networks (CNNs), have undergone extensive scrutiny in the context of food classification because of their exceptional feature extraction capabilities.

Similarly, ensemble-based learning approaches have exhibited great potential for achieving effective supervised classification.

We suggest an innovative approach to improve the effectiveness of deep learning-based food classification.

Our proposal involves a novel deep learning ensemble framework that draws inspiration from the fusion of deep learning models with ensemble learning based on random subspaces. The random subspaces play a role in diversifying the ensemble system in a straightforward but impactful way. Moreover, to enhance the classification accuracy even more, we explore transfer learning, employing the migration of acquired weights from a single classifier to another (namely, CNNs). This approach expedites the process of learning.

Results from experiments conducted using well-established food datasets illustrate that the suggested deep learning ensemble system delivers competitive performance compared to state-of-the-art techniques, as evidenced by its classification accuracy.

The amalgamation of deep learning and ensemble learning holds substantial promise for dependable food categorization.

Indoor space layout planning and design involves sensitive and confidential information. To enhance the security and confidentiality of such data, the study introduces an advanced image encryption algorithm. This algorithm is based on simultaneous chaotic systems and bit plane permutation diffusion, aiming to provide a more secure and reliable approach to indoor space layout design.

The study proposes an image encryption algorithm that incorporates simultaneous chaotic systems and bit plane permutation diffusion. This algorithm is then applied to the process of indoor space layout planning and design. Comparative analysis is conducted to evaluate the performance of the proposed algorithm against other existing methods. Additionally, a comparative testing of indoor space layout planning and design methods is carried out to assess the overall effectiveness of the research method.

Through the algorithm comparison test, information entropy, adjacent pixel distribution and response time were selected as evaluation indexes. The results demonstrated that the improved image encryption algorithm exhibited superior performance in terms of information entropy (with average information entropy of 7.9990), anti-noise attack capability (with PSNR value of 37.58db), and anti-differential attack capability (with NPCR and UACI values of 99.6% and 33.5%) when compared to the benchmark algorithm. In the actual application effect test, the study selected space utilization, functionality, security, ease of use, confidentiality, flexibility and other evaluation indicators. A comparative analysis of the actual application effects of various interior design projects revealed that the interior space layout planning and design method proposed in the study exhibited notable superiority over the comparison method across all indicators. In particular, it showed overall advantages in space utilization (92.5% in modern apartment design), functionality score (9.5 in future living experience museum design), and safety assessment.

The above key results demonstrate that the improved image encryption algorithm and the designed indoor space layout planning method have substantial practical applications and are expected to enhance security and confidentiality in the field of indoor space layout planning, thereby providing users with a more optimal experience.

“Artificial Intelligence will revolutionize our lives” is a phrase frequently echoed. The influence of Artificial Intelligence (AI) and Machine Learning (ML) extends across various aspects of our daily lives, encompassing health, education, economics, the environment, and more.

A particularly formidable challenge lies in decision support, especially in critical scenarios such as natural disaster management, where artificial intelligence significantly amplifies its ongoing capacity to assist in making optimal decisions. In the realm of disaster management, the primary focus often centers on preventing or mitigating the impact of disasters. Consequently, it becomes imperative to anticipate their occurrence in terms of both time and location, enabling the effective implementation of necessary strategies and measures. In our research, we propose a disaster forecasting framework based on a Multi-Layer Perceptron (MLP) empowered by the Particle Swarm Optimization (PSO) algorithm. The PSO-MLP is further fortified by the incorporation of the Artificial Bee Colony (ABC) algorithm for feature selection, pinpointing the most critical elements. Subsequently, we employ the LIME (Local Interpretable Model-agnostic Explanations) model, a component of eXplainable Artificial Intelligence (XAI). This comprehensive approach aims to assist managers and decision-makers in comprehending the factors influencing the determination of the occurrence of such disasters and increases the performance of the PSO-MLP model. The approach, specifically applied to predict snow avalanches, has yielded impressive results.

The obtained accuracy of 0.92 and an AUC of 0.94 demonstrate the effectiveness of the proposed framework. In comparison, the prediction precision achieved through an SVM is 0.75, while the RF classifier yields 0.86, and XGBoost reaches 0.77. Notably, the precision is further enhanced to 0.81 when utilizing XGBoost optimized by the grid-search.

These results highlight the superior performance of the proposed methodology, showcasing its potential for accurate and reliable snow avalanche predictions compared to other established models.

In today’s ever-changing world, military forces face significant challenges in maintaining situational awareness and responding swiftly to emerging threats. Traditional aerial surveillance often fails to give timely and thorough intelligence over large areas. Limited coverage, mistakes, and difficulty noticing small changes on the ground hinder military operations. To address these problems, this paper introduces the development of a deep learning-based web application named “SkyViewSentinel”, a solution tailored specifically for military aerial surveillance.

The application framework i.e., SkyViewSentinel has been developed through multiple stages i.e., (i) pre-process the Xview overhead Satellite imagery dataset using Ground Truth refinement and image partitioning method, (ii) employed a SOTA deep model i.e., YOLOv8 as a baseline architecture for the research problem and assessed the performance on experimental dataset, (iii) a series of rigorous experiments have been conducted using deep model and obtained results are reported. (iv) Finally, the trained model has been seamlessly integrated into the web application and develops a comprehensive web-based object detection application. The developed application detects military-based objects from real-time satellite images.

The developed application has shown promising results in identifying military objects from satellite images, outperforming other contemporary methods. The designed framework has achieved an overall mAP score of 0.315 for all nine classes of military-based objects. For certain specific classes, detection accuracy exceeds 70%, demonstrating the robustness of the framework.

The designed web application enables users to detect military-based objects in the region provided by the user. By harnessing the power of satellite object recognition technology, SkyViewSentinel provides a new way to monitor and understand activities in operational areas.