ArrayPub Date : 2025-04-26DOI: 10.1016/j.array.2025.100400
Mostafa Mansour , Ahmed Abdelsalam , Ari Happonen , Jari Porras , Esa Rahtu
{"title":"UDGS-SLAM: UniDepth Assisted Gaussian Splatting for Monocular SLAM","authors":"Mostafa Mansour , Ahmed Abdelsalam , Ari Happonen , Jari Porras , Esa Rahtu","doi":"10.1016/j.array.2025.100400","DOIUrl":"10.1016/j.array.2025.100400","url":null,"abstract":"<div><div>Recent advancements in monocular neural depth estimation, particularly those achieved by the UniDepth network, have prompted the investigation of integrating UniDepth within a Gaussian splatting framework for monocular SLAM. This study presents UDGS-SLAM, a novel approach that eliminates the necessity of RGB-D sensors for depth estimation within Gaussian splatting framework. UDGS-SLAM employs statistical filtering to ensure local consistency of the estimated depth and jointly optimizes camera trajectory and Gaussian scene representation parameters. The proposed method achieves high-fidelity rendered images and low ATE-RMSE of the camera trajectory. The performance of UDGS-SLAM is rigorously evaluated using the TUM RGB-D dataset and benchmarked against several baseline methods, demonstrating superior performance across various scenarios. Additionally, an ablation study is conducted to validate design choices and investigate the impact of different network backbone encoders on system performance.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100400"},"PeriodicalIF":2.3,"publicationDate":"2025-04-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143890783","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-10DOI: 10.1016/j.array.2025.100389
Md. Shahriar Hossain Apu, Samsuddin Ahmed, Md. Toukir Ahmed
{"title":"Smart system for real time monitoring and diagnosis of dengue surfaces in Bangladesh","authors":"Md. Shahriar Hossain Apu, Samsuddin Ahmed, Md. Toukir Ahmed","doi":"10.1016/j.array.2025.100389","DOIUrl":"10.1016/j.array.2025.100389","url":null,"abstract":"<div><div>Efficient vector management techniques hold great importance according to the World Health Organization (WHO) as a foundation to reduce and maintain the decline of vector-born diseases. Health surveillance teams operating in malaria and dengue and Zika and Chikungunya endemic areas can effectively use drone or unmanned aerial vehicles (UAVs) as technology to detect and eradicate mosquito breeding sites. UAVs enable users to obtain detailed aerial photographs and monitor locations throughout time and geographic areas. The process of vector control intervention analysis through manual image inspection requires extensive labor efforts and takes significant amounts of time. This research presents a methodology to automatically detect mosquito breeding areas in aerial drone images. Leveraging a CBAM-enhanced YOLOv9 object detection framework, we present a UAV-based strategy for dengue surface monitoring, achieving an impressive mean Average Precision (mAP) of 99.5% for mAP50, 86.4% for mAP50-95, 94% for Intersection over Union (IoU), 45 FPS, 98% precision, and 90% recall. The integration of the Convolutional Block Attention Module (CBAM) enhances the model’s feature extraction capabilities, improving its focus on critical regions in the images. Robust performance was ensured by the consistent achievement of these outcomes in a variety of operational and environmental contexts, including urban and rural locations. To confirm the model’s practicality, more tests under various circumstances will be carried out. This deep learning approach facilitates targeted and timely vector control interventions, leveraging drone-based surveillance to combat the spread of vector-borne diseases efficiently.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100389"},"PeriodicalIF":2.3,"publicationDate":"2025-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143823929","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-09DOI: 10.1016/j.array.2025.100394
Youxuan Sun, Yunliang Chen , Xiaohui Huang , Yuewei Wang, Shaoqian Chen, Kangfei Yao, Ao Yang
{"title":"DGTM: Deriving Graph from transformer with Mamba for panoptic scene graph generation","authors":"Youxuan Sun, Yunliang Chen , Xiaohui Huang , Yuewei Wang, Shaoqian Chen, Kangfei Yao, Ao Yang","doi":"10.1016/j.array.2025.100394","DOIUrl":"10.1016/j.array.2025.100394","url":null,"abstract":"<div><div>Scene Graph Generation (SGG) transforms images into structured graph representations that encapsulate the objects, attributes, and relationships present within objects. Graph models boost visual content understanding and reasoning for image captioning, question answering, and HCI. Panoptic Scene Graph Generation (PSG) enhances the object detection task within scene graph generation by incorporating panoptic segmentation, thereby imposing greater demands on the model’s capacity to comprehend images. Existing approaches often rely on intricate modeling techniques to predict relationships between objects, while neglecting the inherent connections among object queries that are learned through multi-head self-attention in object detectors. This oversight not only leads to a significant increase in parameter count but also complicates model design and hinders transferability. This paper proposes a new single-stage panoptic scene graph generator called DGTM (Deriving Graph from transformer with Mamba). DGTM utilizes the by-products of multi-head self-attention layers in transformers, treats queries and keys as subjects and objects respectively to extract relationship information between objects. By introducing the Mamba module, multi-level and multi-scale feature information is integrated efficiently, empowering the model to better grasp intricate relationships. In addition, a Kolmogorov–Arnold Network (KAN) is incorporated to help the model better distinguish between subjects and objects, enriching feature representation. Experimental results show that DGTM achieves at least 25%, 15%, and 15% improvements in mR@20, mR@50, and mR@100 compared to the baseline, demonstrating notable enhancements in the precision and comprehensiveness of PSG.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100394"},"PeriodicalIF":2.3,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143835246","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-07DOI: 10.1016/j.array.2025.100397
Debendra Muduli , Sourav Parija , Suhani Kumari , Asmaul Hassan , Harendra S. Jangwan , Abu Taha Zamani , Sk. Mohammed Gouse , Banshidhar Majhi , Nikhat Parveen
{"title":"Deep learning-based detection and classification of acute lymphoblastic leukemia with explainable AI techniques","authors":"Debendra Muduli , Sourav Parija , Suhani Kumari , Asmaul Hassan , Harendra S. Jangwan , Abu Taha Zamani , Sk. Mohammed Gouse , Banshidhar Majhi , Nikhat Parveen","doi":"10.1016/j.array.2025.100397","DOIUrl":"10.1016/j.array.2025.100397","url":null,"abstract":"<div><div>Leukemia is identified by an excess of immature white blood cells (WBC) being formed in the bone marrow, leading to cancer. It is divided into two main types: acute, which stems from early cell growth ab- normalities and involves rapid immature cell proliferation, and chronic, which progresses more slowly due to a blockage in the later stages of the cell life cycle. Detecting acute lymphoblastic leukemia (ALL) at an early stage is critical to reducing its associated mortality rate. This study presents an empirical analysis of various pre-trained deep learning models, including VGG16, VGG19, ResNet50, Xception, ResNet152, EfficientNet- B0, NASNetMobile, DenseNet169, DenseNet121, and EfficientNetV2B0, for the detection and classification of ALL. A comprehensive evaluation highlights the effectiveness of deep learning in distinguishing different types of ALL, demonstrating its potential as a reliable diagnostic tool in medical imaging. Additionally, we evaluated the performance of these models using different optimization techniques, including Adadelta, SGD, RMSprop, and Adam, to determine the most effective optimization strategy for improving classifica-tion accuracy. Our results demonstrate that EfficientNet-B0 achieved a classification accuracy of 72 %, while NASNetMobile attained 81 %. Notably, DenseNet121 outperformed these models with an accuracy of 99 %. Furthermore, the remaining seven models VGG16, VGG19, ResNet50, Xception, ResNet152, DenseNet169, and EfficientNetV2B achieved a perfect classification accuracy of 100 %, highlighting their robustness and effectiveness in our experimental setup. To improve the interpretability of the leukemia detection process, explainable AI techniques, including Grad-CAM, Score-CAM, and Grad-CAM++, were integrated to vi-sualize critical regions influencing model predictions. These techniques enhance transparency by providing visual explanations of classification decisions. A detailed comparative analysis was conducted, examining key parameters such as learning rate, optimization algorithms, and the number of training epochs to determine the most effective approach. The study leveraged a publicly available acute lymphoblastic leukemia dataset to ensure comprehensive model evaluation. By offering insights into model performance and interpretability.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100397"},"PeriodicalIF":2.3,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143821374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-07DOI: 10.1016/j.array.2025.100392
Francesco Agnelli, Giuseppe Facchi, Giuliano Grossi, Raffaella Lanzarotti
{"title":"KA-GCN: Kernel-Attentive Graph Convolutional Network for 3D face analysis","authors":"Francesco Agnelli, Giuseppe Facchi, Giuliano Grossi, Raffaella Lanzarotti","doi":"10.1016/j.array.2025.100392","DOIUrl":"10.1016/j.array.2025.100392","url":null,"abstract":"<div><div>Graph Structure Learning (GSL) methods address the limitations of real-world graphs by refining their structure and representation. This allows Graph Neural Networks (GNNs) to be applied to broader unstructured domains such as 3D face analysis. GSL can be considered as the dynamic learning of connection weights within a layer of message passing in a GNN, and particularly in a Graph Convolutional Network (GCN). A significant challenge for GSL methods arises in scenarios with limited availability of large datasets, a common issue in 3D face analysis, particularly in medical applications. This constraint limits the applicability of data-intensive GNN models, such as Graph Transformers, which, despite their effectiveness, require large amounts of training data. To address this limitation, we propose the Kernel-Attentive Graph Convolutional Network (KA-GCN). Our key finding is that integrating kernel-based and attention-based mechanisms to dynamically refine distances and learn the adjacency matrix within a Graph Structure Learning (GSL) framework enhances the model’s adaptability, making it particularly effective for 3D face analysis tasks and delivering strong performance in data-scarce scenarios. Comprehensive experiments on the Facescape, Headspace, and Florence datasets, evaluating age, sexual dimorphism, and emotion, demonstrate that our approach outperforms state-of-the-art models in both effectiveness and robustness, achieving an average accuracy improvement of 2%. The project page is available on GitHub <span><span><sup>1</sup></span></span>.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100392"},"PeriodicalIF":2.3,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143807798","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-05DOI: 10.1016/j.array.2025.100396
Jiahui Chen, Mingrui Wu, Huiwu Huang
{"title":"LDAM: A lightweight dual attention module for optimizing automotive malware classification","authors":"Jiahui Chen, Mingrui Wu, Huiwu Huang","doi":"10.1016/j.array.2025.100396","DOIUrl":"10.1016/j.array.2025.100396","url":null,"abstract":"<div><div>In recent years, electric vehicles have become prime targets for cyberattacks, with attackers exploiting public charging stations, USB ports, and other entry points to implant malware. This can lead to network outages and power disruptions. Traditional rule-based classification methods struggle against malware due to advanced encryption and obfuscation techniques. Thus, innovative classification approaches are urgently needed. Previous research often focused on converting malware binary files into RGB images for family classification, but overlooked the importance of runtime memory data. Inspired by earlier work, this study introduces a new deep neural network feature extraction module, the Lightweight Dual Attention Module (LDAM), for malware image classification. LDAM leverages attention mechanisms to capture both global and detailed features and uses feature fusion to balance these scales. This approach allows the neural network to effectively classify both raw binary and memory images of malware, while maintaining a low number of trainable parameters. By integrating LDAM into EfficientNet, the model achieves classification accuracies of 96.7% on the Malvis dataset and 98.1% on the Dumpware10 dataset, making it suitable for classifying malware on vehicle systems.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100396"},"PeriodicalIF":2.3,"publicationDate":"2025-04-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143821375","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Meta-learning based softmax average of convolutional neural networks using multi-layer perceptron for brain tumour classification","authors":"Irwan Budi Santoso, Shoffin Nahwa Utama, Supriyono","doi":"10.1016/j.array.2025.100398","DOIUrl":"10.1016/j.array.2025.100398","url":null,"abstract":"<div><div>Brain tumour classification using Magnetic Resonance Imaging (MRI) is crucial for medical decision-making. The variability in tumour shape, size, and position poses challenges to classification methods. Convolutional Neural Networks (CNNs) are commonly used due to their proven performance, but their effectiveness diminishes with the high variability of tumour characteristics. This study proposes a meta-learning approach, leveraging the softmax average of multiple CNN models with a Multi-Layer Perceptron (MLP) as the meta-learner. The base-learner models include MobileNetV2, InceptionV3, Xception, DenseNet201, and ResNet50. This approach combines the softmax outputs of these CNN models, capturing their strengths to handle diverse tumour characteristics. The averaged outputs are fed into the MLP for increased classification performance. To evaluate the proposed method, we used several brain MRI image datasets, including Dataset 1 (Thomas Dubail Dataset), Dataset 2 (Mesoud Nickparcar Dataset), and Dataset 3 (Fernando Feltrin Dataset). The test results showed the proposed method's effectiveness in improving classification performance. For Dataset 1, the MLP with one hidden layer (128 neurons) achieved 97.47 % accuracy, improving the base learners' performance by 1.94 %–7.42 %. On Dataset 2, the MLP with 64 neurons reached 99.54 % accuracy, with a 0 %–2.44 % improvement. For Dataset 3, an MLP with two hidden layers (256 and 125 neurons) achieved 98.87 % accuracy, enhancing performance by 0.46 %–5.67 %.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100398"},"PeriodicalIF":2.3,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143821373","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-04DOI: 10.1016/j.array.2025.100390
MD Shahriar Mahmud Bhuiyan , MD AL Rafi , Gourab Nicholas Rodrigues , MD Nazmul Hossain Mir , Adit Ishraq , M.F. Mridha , Jungpil Shin
{"title":"Deep learning for algorithmic trading: A systematic review of predictive models and optimization strategies","authors":"MD Shahriar Mahmud Bhuiyan , MD AL Rafi , Gourab Nicholas Rodrigues , MD Nazmul Hossain Mir , Adit Ishraq , M.F. Mridha , Jungpil Shin","doi":"10.1016/j.array.2025.100390","DOIUrl":"10.1016/j.array.2025.100390","url":null,"abstract":"<div><div>Algorithmic trading has revolutionized financial markets, offering rapid and efficient trade execution. The integration of deep learning (DL) into these systems has further enhanced predictive capabilities, providing sophisticated models that capture complex, non-linear market patterns. This systematic literature review explores recent advancements in the application of DL algorithms to algorithmic trading with a focus on optimizing financial market predictions. We analyze and synthesize the key DL architectures, such as recurrent neural networks (RNN), long short-term memory (LSTM), convolutional neural networks (CNN), and hybrid models, to evaluate their performance in predicting stock prices, volatility, and market trends. The review highlights current challenges, such as data noise, overfitting, and interpretability, while discussing emerging solutions and future research directions. Our findings provide a comprehensive understanding of how DL reshapes algorithmic trading and its potential to improve decision-making processes in volatile financial environments.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100390"},"PeriodicalIF":2.3,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143776563","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-02DOI: 10.1016/j.array.2025.100395
Fareeha Naveed , Adven Masih , Jabar Mahmood , Moeez Ahmed , Aitizaz Ali , Aysha Saddiqa , Mohamed Shabbir Hamza Abdulnabi , Ebenezer Agbozo
{"title":"Sustainable AI for plant disease classification using ResNet18 in few-shot learning","authors":"Fareeha Naveed , Adven Masih , Jabar Mahmood , Moeez Ahmed , Aitizaz Ali , Aysha Saddiqa , Mohamed Shabbir Hamza Abdulnabi , Ebenezer Agbozo","doi":"10.1016/j.array.2025.100395","DOIUrl":"10.1016/j.array.2025.100395","url":null,"abstract":"<div><div>Addressing the critical challenge of reduced crop production caused by plant diseases is essential to safeguard agricultural yield and quality. Conventional methods like visual inspection and laboratory testing are both time-consuming and costly. Although Modern AI-based deep learning techniques are promising, their potential in fields such as plant disease identification often remains unexplored due to the requirement of large and expert-labeled data. To mitigate these challenges, it is imperative to explore sustainable approaches that require minimal data while maintaining high accuracy in classification tasks. This research proposes a novel few-shot learning (FSL) framework employing a minimum sample size of 1 image and a maximum of 10 images per class for the accurate classification of plant diseases. The architecture incorporates a pre-training phase based on transfer learning as a feature extractor, followed by meta-learning using Prototypical Networks (ProtoNets) for class prototype computation and distance-based classification. The study evaluates the effectiveness of the proposed approach on the PlantVillage as well as rice disease datasets, performing comparative analyses among different transfer learning models such as ResNet18, ResNet50, and Vision Transformers in combination with Prototypical Networks under various N-way classification tasks (3-way, 5-way, and 10-way) and support sample (K-shot) settings (K=1, K=5, K=10). The experimental findings indicate that the proposed combination of pretraining through ResNet18 with Prototypical Networks achieved an impressive accuracy of 93% and 75% on PlantVillage. The proposed model’s performance was further evaluated on rice disease data where it achieves the average accuracy of 75%. Specifically, the proposed model demonstrated the ability to classify 10 distinct plant diseases with high accuracy when provided with a suitable sample size per class. The proposed framework offers a substantial advancement in sustainable AI for plant disease recognition by enhancing the model generalization, enabling accurate classification across numerous classes with minimal sample size, and addressing data scarcity in AI-driven agricultural solutions.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100395"},"PeriodicalIF":2.3,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143768059","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
ArrayPub Date : 2025-04-02DOI: 10.1016/j.array.2025.100386
Mohammad Rifat Ahmmad Rashid, Md. AL Ehtesum Korim, Mahamudul Hasan, Md Sawkat Ali, Mohammad Manzurul Islam, Taskeed Jabid, Raihan Ul Islam, Maheen Islam
{"title":"An Ensemble Learning Framework with Explainable AI for interpretable leaf disease detection","authors":"Mohammad Rifat Ahmmad Rashid, Md. AL Ehtesum Korim, Mahamudul Hasan, Md Sawkat Ali, Mohammad Manzurul Islam, Taskeed Jabid, Raihan Ul Islam, Maheen Islam","doi":"10.1016/j.array.2025.100386","DOIUrl":"10.1016/j.array.2025.100386","url":null,"abstract":"<div><div>The early and accurate detection of plant diseases is critical for sustainable agriculture, ensuring crop health, reducing losses, and supporting food security. To address this challenge, we present an Ensemble Learning Framework with Explainable AI (XAI) tailored to disease detection, using cucumber leaf diagnosis as a key use case. In this study, we experimented with a dataset comprising 6,400 images capturing six prevalent cucumber leaf diseases – Gummy Stem Blight, Downy Mildew, Anthracnose, Bacterial Wilt, Belly Rot, and Pythium Fruit Rot – alongside two healthy categories. Prior to training, the images underwent preprocessing steps such as resizing, rescaling, and data augmentation (through random rotations, flips, zooms, and contrast adjustments) to enhance model generalization. The proposed framework unites multiple architectures – CNN, DenseNet121, EfficientNetB0, InceptionV3, MobileNetV2, ResNet50, and Xception – into an ensemble that attained overall accuracy of 99%, alongside high recall and F1-scores. Individual models demonstrated accuracy ranging from 88.71% to 99%, underscoring the robustness of the ensemble. Integrating XAI methods further ensures interpretable outputs, granting valuable insights into the decision-making process and heightening transparency for researchers and agronomists. The findings confirm that transfer learning, model ensembling, and interpretability methods significantly enhance classification performance, especially in cases of limited data, offering a scalable solution for improved disease management in agriculture. Additionally, the framework is scalable for real-world deployment by enabling real-time disease monitoring on edge devices (e.g., Raspberry Pi, IoT systems), seamless integration with smart farming platforms, and continuous learning for adaptive crop management.</div></div>","PeriodicalId":8417,"journal":{"name":"Array","volume":"26 ","pages":"Article 100386"},"PeriodicalIF":2.3,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143759583","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}