Deep Learning Based Multiclass Detection of Corn Leaf Diseases Using a Convolutional Neural Network

Deep Learning Based Multiclass Detection of Corn Leaf Diseases Using a Convolutional Neural Network | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Deep Learning Based Multiclass Detection of Corn Leaf Diseases Using a Convolutional Neural Network Ismail Ismail, Nursakti Nursakti, Lut Faizal, Ibrahim Ibrahim This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7872919/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The research develop an accurate and efficient method for detecting multiple corn leaf diseases to support sustainable agricultural practices in Soppeng Regency, Indonesia. The goal is to design a Convolutional Neural Network (CNN) model capable of classifying corn leaf diseases, including rust, blight, and gray leaf spot, using high-resolution image data. The research employed a balanced dataset sourced from open-access repositories, followed by preprocessing, data augmentation, and CNN model optimization. The model’s performance was evaluated using accuracy, precision, recall, and F1-score to ensure comprehensive assessment. Experimental results show that the proposed CNN achieved high accuracy across all disease classes, with strong per-class metrics, indicating robust performance in distinguishing visually similar symptoms. The classification results with the Convolutional Neural Network algorithm have 95% training data accuracy and 93% test data accuracy in detecting leaf diseases in corn plants. The findings contribute to agricultural technology by offering a scalable and field-deployable disease detection system that can be integrated into mobile or edge-based platforms. Limitations include reliance on publicly available datasets, which may not fully capture the variability of local field conditions. The research concludes that the proposed CNN model can significantly enhance early disease detection, reduce dependency on manual inspections, and support precision agriculture. Future research should focus on expanding the dataset with locally captured images, incorporating real-time image acquisition, and optimizing the model for deployment in low-resource environments to improve adaptability and reliability. Convolutional Neural Network Corn Leaf Disease Detection Image Classification Precision Agriculture Deep Learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. Introduction Corn is a food plant that occupies the third position in the world after rice and wheat, and ranks second in Indonesia after rice [ 1 ]. Corn contains carbohydrates, which have the potential to be developed as functional food products because corn contains essential amino acids, minerals, dietary fiber, essential fatty acids, and others needed by the body. Increasing consumer demand, of course, will determine the production quantity for corn cultivation [ 2 ]. According to research, there are various aspects that affect production in corn cultivation, namely seeds, land area, fertilization accuracy, and labor. Fertilization aims to improve crop quality and prevent pests or corn diseases. Pests can cause crop yields to be less than optimal, and even harvest failure can occur, so maximizing the amount of corn production requires the right steps or cultivation to anticipate pest attacks that can cause various diseases in corn plants [ 3 ]. Corn plant diseases can be identified manually with the naked eye, because the color of the corn leaves will change each time it is affected by the disease. However, the color of the leaves is rather difficult to distinguish because the color is almost the same, apart from that, of course, it will take a long time if the plants are identified in large quantities, especially since each individual has a different color assessment [ 4 ]. Identifying this disease requires an expert, because the disease's lethal characteristics and lack of similarities can give the wrong impression. One method that can be used to classify leaf diseases is using image datasets, namely deep learning[ 5 ]. Deep Learning is a branch of broader machine learning using artificial neural networks. Deep learning architectures can be artificial neural networks, recurrent neural networks, and convolutional neural networks that have been applied to areas such as pattern recognition, speech recognition, natural language processing, audio recognition, social network filtering, machine translation, bioinformatics, medical image analysis, and other fields with satisfactory results [ 6 ]. There has been a lot of deep learning development, such as CNN, which has been widely researched in terms of making models to get high accuracy values. This CNN is a refinement of the multi-layer perceptron or MLP algorithm, built to manage 2-dimensional data. CNN belongs to the deep neural network category because it is a fairly good network and has been widely applied for processing image data. In the CNN Algorithm, there is an arrangement of text into a matrix, with each row representing a word embedding, a word, or a character [ 7 ]. By utilizing the Convolutional Neural Network (CNN) method as one of the Machine Learning technologies, the problem of classifying the image of corn plant leaves will be easier to solve [ 8 ]. Research has been carried out using artificial neural network methods based on backpropagation algorithms, with an accuracy rate of 90% and an error of 10% [ 9 ]. The next researcher detected diseases in corn plants based on image processing using the color moments and GLCM methods. This method achieves an accuracy rate of 89.375% [ 10 ]. Another study with a different object, namely the detection of disease in potatoes using the CNN method, resulted in a training accuracy of 95% with a validation accuracy of 94% [ 11 ]. While prior studies have applied CNN and other machine learning models for plant disease detection, most have focused on binary classification or limited crop types. Moreover, many works lack validation using robust performance metrics or rely on manually engineered features [ 12 ]. This study presents a comprehensive multiclass classification of corn leaf diseases using an end-to-end CNN model trained from scratch on a curated public dataset [ 13 ]. The novelty of this work lies in: 1) Application of a custom CNN architecture specifically tailored for corn leaf imagery rather than relying on generic or pre-trained models. 2) Multiclass classification across four disease categories, including subtle differentiations between Blight, Rust, and Gray Leaf Spot, which are visually similar. 3) Quantitative evaluation using a full suite of metrics (precision, recall, F1-score, confusion matrix) to ensure class-level performance is well understood. 4) Implementation of a scalable preprocessing pipeline, enabling reproducibility and real-world deployment in smart agriculture systems. This research contributes to the development of automated, accurate, and scalable solutions for disease detection in crops, particularly in contexts with limited access to expert agronomists 2. Related work Several studies have explored the application of machine learning and deep learning techniques to plant disease classification using leaf imagery. Traditional image processing approaches, such as those based on color moment and texture features, have demonstrated moderate success[ 14 ]. For example, employed the Color Moment and Gray-Level Co-occurrence Matrix (GLCM) methods to detect rubber plant diseases, achieving an accuracy of 89.37%. However, these approaches often rely on handcrafted features, which may not generalize well across different plant species and lighting conditions. With the emergence of deep learning, Convolutional Neural Networks (CNNs) have become a dominant approach for image-based classification tasks[ 15 ]. Applied a pre-trained deep learning model to classify corn leaf diseases and achieved an accuracy of 90%, indicating the potential of CNNs for agricultural applications. Similarly, developed a CNN model for detecting diseases in potato leaves, reporting a training accuracy of 95% and a validation accuracy of 94% [ 16 ]. While promising, these studies were either limited to binary classification or involved crops other than corn, which poses a challenge for generalization in corn-specific pathology [ 17 ]. Explored the VGG16 pre-trained CNN architecture for grape leaf disease classification, which demonstrated high accuracy. However, pre-trained networks such as VGG16, while effective, often contain large parameter sizes and are not optimized for lightweight deployment in agricultural field conditions. Additionally, their reliance on transfer learning may introduce bias when applied to non-natural image datasets such as plant leaves. More recent efforts have aimed at developing mobile-based or lightweight CNN variants tailored for specific plants [ 18 ]. Introduced a CNN-Squeeze Net-based application for detecting papaya diseases, showing that custom architectures can outperform generic models in domain-specific tasks. Despite this progress, most existing models are either trained on small, imbalanced datasets or lack rigorous evaluation through per-class metrics such as precision, recall, and F1-score. In the context of corn plant disease detection, a significant gap remains in the development of a CNN model specifically trained and evaluated on a multiclass corn leaf disease dataset. Moreover, the use of balanced, high-resolution image data from open-source repositories like Kaggle has not been fully leveraged to build robust models tailored to the characteristics of corn leaves, which often display overlapping visual symptoms such as rust, blight, and gray leaf spot [ 19 ]. Recent studies increasingly show that Vision Transformers (ViT) and hybrid CNN-Transformer architectures can outperform or at least match classic CNN backbones for crop disease recognition—especially on complex, in-field images. Hybrid pipelines that fuse local CNN features with ViT’s global context have improved transferability and feature discrimination under variable illumination and background clutter, narrowing the lab-to-field gap (e.g., MViT + CNN frameworks)[ 20 ]. Comprehensive 2025 surveys of maize leaf disease detection echo this trend, reporting stronger benchmarks from Swin/ViT variants and CNN-ViT hybrids compared with pure CNNs, while also highlighting dataset curation and augmentation as performance drivers[ 21 ]. Recent transformer-centric solutions (e.g., Swin-based or ViT-enhanced classifiers) continue to set competitive results on PlantVillage-style corpora and mixed field datasets, underscoring the growing role of attention mechanisms in modeling subtle visual symptoms on maize leaves [ 22 ]. In parallel, there’s a clear push toward lightweight models suitable for mobile and edge deployment so that diagnosis can run on-device without cloud connectivity. 2025 work proposes compact MobileNet-style variants (e.g., RTR_Lite_MobileNet) that retain accuracy while cutting parameters and latency, making them attractive for farm-side use in low-resource settings[ 23 ]. Similar efforts in maize-specific classifiers emphasize interpretability and efficiency to aid extension workers and farmers, balancing accuracy with model size and explainability. More broadly, edge-computing studies in plant phenotyping argue that optimizing architectures and memory footprints is essential for practical, field-ready systems—reinforcing the relevance of lightweight CNNs and quantization/pruning for deployment [ 24 ]. Finally, to improve robustness and generalization beyond curated datasets, recent work explores domain adaptation, few-shot learning, and real-time detection. Unsupervised domain adaptation methods mitigate performance drops when moving from lab images to diverse field conditions with shifting backgrounds and lighting[ 25 ]. Cross-domain few-shot frameworks enable fast adaptation to unseen farms or newly emerging maize diseases with very limited labeled samples, which is crucial where expert annotation is scarce. Alongside classification, object-detection pipelines tailored to maize leaves (e.g., YOLO-style models with multi-scale kernels) enable real-time localization of symptomatic regions directly in the field, supporting precision scouting and timely intervention [ 26 ]. This study addresses the aforementioned gaps by proposing a custom CNN model trained from scratch on a curated and balanced dataset of corn leaf images. It employs a standardized preprocessing pipeline and conducts a comprehensive evaluation using class-wise precision, recall, and F1-score. Unlike prior works, the proposed approach focuses specifically on multiclass disease classification in corn , which is crucial for precision agriculture applications in low-resource settings. 3. Research Method This research had several stages in perfecting the grouping or classification stage of corn leaf disease, including problem identification, literature study, dataset collection, dataset processing, CNN, and results evaluation. Solving the problem using the Convolutional Neural Network (CNN) method in identifying corn plant leaves [ 27 ]. The stages of the research methodology were: 1) Problem Identification. At this stage, data collection and problem analysis were carried out. 2) Literature study, After collecting the dataset used in data processing, data processing was carried out using the Convolution Neural Network algorithm. 3) Evaluation of Results, After processing the data using the Convolution Neural Network algorithm, the results were then evaluated. [ 28 ]. The following was the methodology in this study, which is described in the form of a diagram flow as shown in Fig. 1 . Collection of Datasets The data used in this study was the corn leaf dataset obtained from the website with the URL: https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset accessed on Table 1 , 11 February 2025. Table 1 The corn leaf datasets [ 29 ] No Class Total 1 Blight 1.146 2 Common Rust 1.306 3 Gray Leaf Spot 574 4 Healthy 1.162 Total 4.188 Types of leaf diseases in corn plants include blight, common rust, grey leaf spot, and healthy. Blight-type corn plant disease is an attack that occurs early in growth, causing the plant to wither and die. Common rust is a leaf rust disease on corn caused by a fungus. Gray leaf spot is an infectious disease that can be carried by wind or rain splashes and can cause the first infection in corn plants. Healthy is a disease characterized by a change in leaf color from pale green to yellow, accompanied by regular linear lines or parallel wavy edges. The following is Fig. 2 , of a disease on corn leaves [ 30 ]. Source: https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset Dataset Processing Data preprocessing serves as a crucial initial step in preparing image datasets for classification tasks, particularly in deep learning applications. In this study, the preprocessing phase involved partitioning the dataset into distinct subsets for training and testing, following an 80:20 ratio to ensure an adequate representation of all classes during model learning and evaluation. Additionally, to maintain consistency and optimize computational efficiency, all images were resized to a uniform resolution of 150 × 150 pixels. This resizing not only standardizes the input dimensions across the dataset but also ensures compatibility with the input layer requirements of the Convolutional Neural Network (CNN) model. Normalization techniques were also applied to scale pixel values within a range suitable for neural network training, thereby improving convergence during optimization [ 31 ]. These preprocessing steps are essential for minimizing data inconsistencies, reducing noise, and enhancing the model’s ability to learn discriminative features effectively. Convolutional Neural Network (CNN) A convolutional neural network is a type of neural network that can be used in image data. CNN is also commonly used to find and identify objects in the form of images. CNN is used in solving the problem of mapping corn leaf area, so that the data can be used to solve problems in classifying corn diseases [ 32 ]. CNN is used to group images and is also used to detect objects in images [ 33 ]. CNN has a series of 3D neurons (width, height, and intensity) which are the shape of the layers, while the intensity or depth is the number of layers. There are generally 2 (two) layers in CNN, namely: feature learning and classification, as shown in Fig. 3 , [ 34 ]. The layers in the Convolutional Neural Network are as follows: 1. Layer feature learning /image feature extraction layer, composed of several layers, and each layer is composed of various neurons that are connected to the local location of the previous layer. The first type of layer is the convolution layer, and the second layer is the pooling layer, which is alternately located. 2. Layer classification/classification, systematically from various layers, with each layer being composed of neurons that are fully connected to other layers. The output of this layer is in the form of class scoring for grouping [ 35 ]. The research object was carried out to identify disease problems in the leaves of corn plants using the CNN algorithm with its architecture [ 36 ]. A convolutional neural network architecture is commonly used in image data. CNN is not much different from neural networks in general. CNN also consists of neurons that have weights, biases, and activation functions. The proposed CNN architecture for identifying corn leaf diseases is shown in Fig. 4 , [ 37 ]: Confusion Matrix The form of the method for evaluating classification is commonly called the confusion matrix and is usually used for calculating recall, error rate, precision, and accuracy. Recall is the ability of the system to determine relevant items, which is defined as the percentage of documents that apply to the query. Accuracy is a comparison where the overall number of errors is correctly identified, and the things that are identified incorrectly are divided by the total number of all cases. Precision evaluates whether the system is capable of detecting the most relevant rating [ 34 ]. The confusion matrix is ​​a table that states the classification of the correct number of test data and the incorrect number of test data [ 38 ]. The following Table 2 is a multiclass confusion matrix. Table 2 Multiclass Confusion Matrix (MCM)[ 38 ] PREDICTION POSITIF NEGATIF NEUTRAL ACTUAL POSITIF TPos FPosNeg FPosNet NEGATIF FNegPos TNeg FNegNet NEUTRAL FNetpos FNetNeg TNet Meanwhile, to calculate the accuracy evaluated in terms of overall accuracy (OA), precision, recall, and F1-score can be calculated using the equations (1), (2), (3), and (4) [ 39 ]: OA = \(\:\frac{\text{T}\text{P}\:+\:\text{T}\text{N}}{\text{T}\text{P}\:+\:\text{F}\text{P}\:+\:\text{T}\text{N}\:+\:\text{F}\text{N}}\) (1) Precision = \(\:\frac{\text{T}\text{P}}{\text{T}\text{P}\:+\:\text{F}\text{P}\:}\) (2) Recall = \(\:\frac{\text{T}\text{P}}{\text{T}\text{P}\:+\:\text{F}\text{N}}\) (3) F1 - Score = \(\:\frac{2\:\times\:\:\text{T}\text{P}}{2\:\times\:\:\text{T}\text{P}\:+\:\text{F}\text{P}\:+\:\text{F}\text{N}}\) (4) Here is the Equation Information: OA = Overall accuracy TP = True Positive TN = True Negative FN = False Negative FP = False Positive 4. Results The classification model was trained and evaluated using a dataset of 4,188 corn leaf images, which were divided into training and testing subsets using an 80:20 split. The training set comprised 3,348 images, while the remaining 840 images were used for testing. The proposed Convolutional Neural Network (CNN) architecture was trained for 100 epochs using a batch size of 32, and validation was conducted using a batch size of 8. Figure 5 shows the output of the designed fit model. The training log presented in Fig. 6 illustrates the model's performance over 100 epochs, demonstrating a consistent trend of convergence and generalization. The final training accuracy reached 95.34% , while the corresponding validation accuracy stabilized at 93.93% , indicating that the model achieved high predictive performance with minimal overfitting. Notably, the validation accuracy remained above 92% throughout the final epochs, suggesting robust learning across all classes. Although a slight increase in validation loss was observed near the final iterations—rising to 0.3617 —the model maintained stable validation accuracy, implying resilience against overfitting. The progressive reduction in training loss, culminating at 0.1219 , further confirms the model's ability to learn discriminative features effectively. These results substantiate the model’s capability to generalize well to unseen data, thereby supporting its applicability for real-world classification of corn leaf diseases. Figure 6 illustrates the model accuracy progression over 100 training epochs for both training and validation sets. The training accuracy demonstrates a steady upward trend, ultimately reaching approximately 95% , while the validation accuracy closely follows, stabilizing around 93–94% . The early epochs display notable fluctuations in validation performance, which is typical as the model adjusts its internal parameters; however, the fluctuations gradually decrease, indicating improved stability and generalization. The convergence of both curves after epoch 30, with minimal divergence between training and validation accuracy, suggests that the model effectively learned the data distribution without overfitting. This level of consistency confirms the robustness of the proposed CNN architecture and highlights its potential for accurate classification of corn leaf diseases in real-world applications. Figure 7 presents the training and validation loss curves over 100 epochs, offering insights into the model’s learning dynamics and generalization capability. The training loss exhibits a smooth and steady decline, reaching a final value below 0.15, which indicates that the model effectively minimized error on the training set. In contrast, the validation loss initially follows a similar downward trend but begins to fluctuate after approximately 20 epochs, stabilizing around 0.35. These fluctuations are typical in deep learning models trained on natural image data and may reflect minor variations in batch composition or limited noise robustness. Despite the variability in validation loss, no significant divergence from the training loss is observed, suggesting that overfitting remains well-controlled. The consistently low training loss and moderately stable validation loss confirm that the model achieved a balanced fit and retained its ability to generalize across unseen data, thus validating its reliability for multiclass corn disease classification tasks. Table 3 Classification Report [ 40 ] PREDICTION Support Blight Common Rust Gray Leaf Spot Healthy ACTUAL Blight 211 9 10 0 230 Common Rust 7 252 3 0 262 Gray Leaf Spot 17 3 95 0 115 Healthy 0 2 0 231 233 235 266 108 231 840 Accuracy = (211 + 252 + 95 + 231) / 840 = 0,939 Precision Recal F1-Score Blight 211/235 = 0,897 211/230 = 0,917 (2×0,897×0,917)/ (0,897 + 0,917) = 0,906 Common Rust 252/266 = 0,947 252/262 = 0,961 (2×0,947×0,961)/ (0,947 + 0,961) = 0,953 Gray Leaf Spot 95/108 = 0,879 95/115 = 0,826 (2×0,879×0,826)/ (0,879 + 0,826) = 0,851 Healthy 231/231 = 1 231/233 = 0,991 (2×1 ×0,991)/ (1 + 0,991) = 0,995 Table 3 presents the detailed classification report, highlighting the model’s predictive performance across four categories of corn leaf health: Blight, Common Rust, Gray Leaf Spot, and Healthy. The overall accuracy of the model reached 93.9%, demonstrating strong generalization capability. Notably, the Healthy class achieved a perfect precision score of 1.00, with a recall of 0.991, resulting in an exceptional F1-score of 0.995, indicating that the model identified healthy leaves with near-perfect reliability. The Common Rust class also performed robustly, with a precision of 0.947, a recall of 0.961, and an F1-score of 0.953, suggesting the model's effectiveness in recognizing this disease despite its visual similarity to other conditions. Although Gray Leaf Spot exhibited slightly lower metrics (F1-score: 0.851), the results remain acceptable given the class’s lower sample size and inter-class feature overlap. Blight showed balanced performance with a precision of 0.897 and a recall of 0.917. These findings indicate that the model performs well across all classes, particularly in high-confidence identification of healthy and rust-infected leaves, making it a reliable tool for automated disease diagnosis in corn cultivation. This study uses the sklearn Python module to visualize the Confusion Matrix so that it can provide information in the form of system detection results for each class tested based on image data testing. The results of system detection can be seen in the following figure. Figure 8 is the result of a visualization of the 4×4 confusion matrix table, which displays the prediction results of the system. The explanation for each class is as follows: 1. Blight class: From a total of 230 test data, the system recognizes 211 images according to their class, 9 data points that should be classified as blight class but are detected as common rust class, and 10 data points are detected as grey leaf spot class. 2. Common rust class: From a total of 262 test data, the system recognizes 252 images according to their class, 7 data points that should be of the common rust blight class but are detected as the blight class, and 3 data points are detected as the grey leaf spot class. 3. Gray leaf spot class: From a total of 115 test data, the system recognizes 95 images according to their class, 17 data points that should be of the grey leaf spot class but are detected as blight class, and 3 data points are detected as common rust class. 4. Healthy rust class: From a total of 233 test data, the system recognizes 231 images according to their class, and 2 data points that do not match their class because they are detected as common rust class In addition to visualizing the test results in the form of a confusion matrix table, the system is also tested for performance by calculating the accuracy, precision, recall, and F1-score values. The performance test results can be seen in the following Fig. 9 : Accuracy, precision, recall, and F1-score values ​​are shown in Fig. 9 as a result of testing system performance using the scikit learn library. In testing 230 blight class images, a precision value of 0.90 was obtained, a recall value of 0.92, and an F1-score of 0.91. The precision value for the common rust class is 0.95, the recall value is 0.96, and the F1-score is 0.95. The precision value for the gray leaf spot class is 0.88, the recall value is 0.83, and the F1-score is 0.85. In the healthy class, the precision value is 1, the recall is 0.99, and the F1-score is 1. The average value of the accuracy of the system for detecting corn leaf disease is 94% 5. Discussion This research was designed to address the need for accurate, efficient, and automated detection of corn leaf diseases, particularly in the context of smallholder farming in Soppeng Regency. The primary objective was to develop and evaluate a Convolutional Neural Network (CNN) model capable of classifying multiple corn leaf diseases—such as rust, blight, and gray leaf spot—based on image data. By leveraging high-resolution, balanced datasets from open-source repositories, the research aimed to overcome limitations found in previous studies, such as reliance on binary classification, small or imbalanced datasets, and lack of field-ready deployment strategies. The experimental results demonstrate that the proposed CNN model achieved high classification accuracy across all targeted corn leaf disease categories, with robust performance in both training and validation phases. The model consistently maintained strong per-class metrics, including precision, recall, and F1-score, indicating its reliability in distinguishing between diseases with visually overlapping symptoms. This outcome confirms that a properly optimized CNN architecture, trained on a well-curated dataset, can effectively handle the complexities of corn pathology in diverse field conditions. The novelty of this research lies in its integration of a CNN architecture specifically trained for multiclass corn disease detection using a balanced and high-quality dataset. Unlike many earlier works that focused on single-disease detection or crops other than corn, this study targets multiple diseases simultaneously while mitigating class imbalance through data preprocessing techniques. Furthermore, the approach emphasizes high-resolution image analysis, enabling finer feature extraction and improved discrimination between diseases with similar visual characteristics. This research contributes to the agricultural technology domain by providing an end-to-end framework for automated disease detection that is adaptable to practical farming environments. The proposed CNN model can be integrated into mobile or edge computing platforms, enabling farmers and agricultural extension workers to diagnose plant diseases rapidly in the field. In addition, the study offers a reproducible methodology for dataset preparation, model training, and evaluation, which can serve as a reference for future research in smart agriculture and precision farming. From a sustainability perspective, the system developed in this study has the potential to reduce dependency on expert manual inspections, thereby lowering operational costs and enabling faster disease management interventions. By facilitating early and accurate detection, it can help prevent severe crop losses, reduce pesticide misuse, and promote more targeted and environmentally friendly agricultural practices. The adaptability of the framework to other crops and regions ensures that the impact of this research can extend beyond corn cultivation in Soppeng Regency, contributing to broader goals in sustainable food production. The proposed CNN-based system demonstrates a reliable, accurate, and efficient solution for detecting multiple corn leaf diseases. The results validate the research objectives and highlight the model’s potential for real-world agricultural deployment. The novelty in dataset utilization, the methodological rigor, and the practical adaptability make this work a valuable contribution to both academic research and applied farming practices. Future work could focus on integrating real-time image acquisition, expanding the dataset with locally sourced images, and optimizing the model for even greater efficiency in low-resource settings. Based on the research findings, the proposed Convolutional Neural Network (CNN) model has proven effective in accurately classifying multiple corn leaf diseases, including rust, blight, and gray leaf spot, using high-resolution and balanced image data. The model achieved high accuracy and strong per-class performance metrics, demonstrating its reliability in distinguishing diseases with similar visual characteristics. These results confirm that a well-trained CNN architecture, supported by quality datasets, can overcome the limitations of traditional methods and earlier studies. Furthermore, the model’s potential for integration into mobile or edge-based platforms makes it a practical tool for real-time disease detection in the field, supporting timely decision-making and contributing to sustainable agricultural practices. 6. Conclusion This research succeeded in implementing machine learning using the Convolutional Neural Network (CNN) algorithm with the sklearn Python library to produce a good level of training accuracy. Presents a high-performing Convolutional Neural Network (CNN) model designed specifically for the multiclass classification of corn leaf diseases using image data. The model was trained and validated on a curated dataset of 4,188 corn leaf images, comprising four classes: Blight, Common Rust, Gray Leaf Spot, and Healthy. The final model achieved an overall accuracy of 93.9%, with class-specific F1-scores ranging from 0.851 to 0.995. The Healthy and Common Rust classes exhibited the highest performance, indicating the model’s strong ability to identify clear visual symptoms. Despite some challenges in distinguishing Gray Leaf Spot due to its visual similarity to other classes, the results demonstrate the model’s robustness and generalizability. The use of a custom-built CNN architecture, combined with a systematic preprocessing pipeline, significantly enhanced learning efficiency and prediction consistency. These findings validate the feasibility of applying deep learning for automated, precise, and scalable disease diagnosis in corn cultivation, offering valuable potential for integration into smart agriculture systems, especially in low-resource settings. Declarations Conflict of Interest The authors hereby declare that there are no known financial, personal, or professional conflicts of interest that could have influenced the conduct, analysis, or reporting of this research. All stages of the study, including data collection, model development, interpretation of results, and manuscript preparation, were carried out independently and without any undue influence from external parties. The funding provided by Lamappapoleonro University was granted to support academic research and did not affect the objectivity, integrity, or transparency of the work presented in this article. As such, the authors affirm that the findings and conclusions are solely the result of scholarly investigation. Ethics approval Ethics statement Not applicable because this research did not involve human participants, animals, or personally identifiable data. All datasets used were obtained from open access repositories and are freely available for academic and research purposes. This study obtained consent from all participants prior to data collection, and confidentiality was maintained throughout the research. Consent to participate All authors have read and approved the final version of this manuscript and consent to its submission for publication. Each author confirms their active participation in the research, data analysis, and manuscript preparation in accordance with ethical publication standards. The authors collectively agree to be accountable for all aspects of the work, ensuring the integrity and accuracy of the published article. Consent to publish This research did not involve human participants, animals, or personally identifiable data. Therefore, consent to publish the research results presented in this article is not required. Clinical Trial Number This study did not involve any clinical trials; therefore, a clinical trial registration number is not applicable. The research was conducted using non-clinical methods and did not include any interventions or procedures involving human participants or patients. All data were collected and analyzed in accordance with ethical research standards without the need for clinical trial registration. Author Contribution All authors contributed significantly to the development of this research. **Ismail** formulated the study concept, designed the CNN model architecture, and led the data analysis and interpretation. He also prepared the initial draft of the manuscript. **Nursakti** was involved in the preprocessing of the dataset. **L Faizal** performed the performance evaluation and refinement of the deep learning framework. **Ibrahim** provided critical revisions to the manuscript and contributed to the discussion and conclusion sections. All authors reviewed and approved the final version of the manuscript and agreed to be accountable for all aspects of the work to ensure its accuracy and integrity. Acknowledgement The authors would like to express their sincere gratitude to the Rector of Lamappapoleonro University for the invaluable support and funding provided for this research. The opportunity and resources extended by the university leadership have played a pivotal role in the successful completion of this study. This support reflects the institution's strong commitment to advancing research and innovation in the field of agricultural technology and artificial intelligence Data Availability The data used in this study comes from a publicly available and curated image repository. The dataset was obtained from the “Corn or Maize Leaf Disease Dataset” . The data source can be seen in the Kaggle – https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset References Rozi F, et al. Indonesian market demand patterns for food commodity sources of carbohydrates in facing the global food crisis. Heliyon. 2023;9(6):e16809. 10.1016/j.heliyon.2023.e16809 . Yu B-G, Chen X-X, Zhou C-X, Ding T-B, Wang Z-H, Zou C-Q. Nutritional composition of maize grain associated with phosphorus and zinc fertilization. J Food Compos Anal. 2022;114:104775. https://doi.org/10.1016/j.jfca.2022.104775 . Morales H, Perfecto I, Ferguson B. Traditional fertilization and its effect on corn insect populations in the Guatemalan highlands. Agric Ecosyst Environ. 2001;84(2):145–55. https://doi.org/10.1016/S0167-8809(00)00200-0 . Khan SU, Alsuhaibani A, Alabduljabbar A, Almarshad F, Altherwy YN, Akram T. A review on automated plant disease detection: motivation, limitations, challenges, and recent advancements for future research. J King Saud Univ Comput Inf Sci. 2025;37(3):34. 10.1007/s44443-025-00040-3 . George R, et al. Past, present and future of deep plant leaf disease recognition: A survey. Comput Electron Agric. 2025;234:110128. https://doi.org/10.1016/j.compag.2025.110128 . Mienye ID, Swart TG. A Comprehensive Review of Deep Learning: Architectures, Recent Advances, and Applications. Inf. 2024;15(12). 10.3390/info15120755 . Zhao X, Wang L, Zhang Y, Han X, Deveci M, Parmar M. A review of convolutional neural networks in computer vision. Artif Intell Rev. 2024;57(4):99. 10.1007/s10462-024-10721-6 . Rachmad A, Fuad M, Rochman EMS. Convolutional Neural Network-Based Classification Model of Corn Leaf Disease. Math Model Eng Probl. 2023;10(2):530–6. 10.18280/mmep.100220 . Rosadi MI, Lutfi M. Identifikasi Jenis Penyakit Daun Jagung Menggunakan Deep Learning Pre- Trained Model. J Explor IT. 2021;13(36):1–8. Purnomo H, Mudjirahardjo P, Pramono SH. Implementation of CBIR Method for Identification of Corn Disease Using Extreme Learning Machine. Int J Comput Appl Technol Res. 2022;11(02):14–21. 10.7753/ijcatr1102.1002 . Dame TA, Adera GB, Girmaw DW. Deep learning-based potato leaf disease classification and severity assessment. Discov Appl Sci. 2025;7(7):649. 10.1007/s42452-025-07240-6 . Myllyaho L, Raatikainen M, Männistö T, Mikkonen T, Nurminen JK. Systematic literature review of validation methods for AI systems. J Syst Softw. 2021;181:111050. https://doi.org/10.1016/j.jss.2021.111050 . Amin H, Darwish A, Hassanien AE, Soliman M. End-to-End Deep Learning Model for Corn Leaf Disease Classification. IEEE Access. 2022;10:31103–15. 10.1109/ACCESS.2022.3159678 . Mehta AR, Kumar P, Prem G, Aggarwal S, Kumar R. AI-powered Innovations in Agriculture: A Systematic Review on Plant Disease Detection and Classification, Indian J. Agric. Res. , no. Of, pp. 1–11, 2025, 10.18805/ijare.a-6371 Salka TD, Hanafi MB, Rahman SMSAA, Zulperi DBM, Omar Z. Plant leaf disease detection and classification using convolution neural networks model: a review. Artif Intell Rev. 2025;58(10):322. 10.1007/s10462-025-11234-6 . Gülmez B. A Comprehensive Review of Convolutional Neural Networks based Disease Detection Strategies in Potato Agriculture. Potato Res. 2025;68(2):1295–329. 10.1007/s11540-024-09786-1 . Ahmad A, Gamal AE, Saraswat D. Toward Generalization of Deep Learning-Based Plant Disease Identification Under Controlled and Field Conditions. IEEE Access. 2023;11:9042–57. 10.1109/ACCESS.2023.3240100 . Karim MJ, Goni MOF, Nahiduzzaman M, Ahsan M, Haider J, Kowalski M. Enhancing agriculture through real-time grape leaf disease classification via an edge device with a lightweight CNN architecture and Grad-CAM. Sci Rep. 2024;14(1):16022. 10.1038/s41598-024-66989-9 . Kabel SAE-M, El-Shafai W, El-Samie FEA, Mohamed B. Enhanced plant identification and disease diagnosis through SqueezeNet and SVM for smart agriculture applications. Iran J Comput Sci. 2025. 10.1007/s42044-025-00237-9 . Tunio MH, et al. Advancing plant disease classification: A robust and generalized approach with transformer-fused convolution and Wasserstein domain adaptation. Comput Electron Agric. 2024;227:109574. https://doi.org/10.1016/j.compag.2024.109574 . Timilsina S, Sharma S, Kondo S. Advancements in maize leaf disease detection, segmentation and classification: A review. Biosyst Eng. 2025;255:104162. https://doi.org/10.1016/j.biosystemseng.2025.104162 . Liu W, Zhang A. Plant Disease Detection Algorithm Based on Efficient Swin Transformer. Comput Mater Contin. 2025;82(2):3045–68. 10.32604/cmc.2024.058640 . Duhan S, Gulia P, Gill NS, Narwal E. RTR_Lite_MobileNetV2: A lightweight and efficient model for plant disease detection and classification. Curr Plant Biol. 2025;42:100459. https://doi.org/10.1016/j.cpb.2025.100459 . Khan AT, Jensen SM, Khan AR, Li S. Plant disease detection model for edge computing devices, Front. Plant Sci. , vol. 14, no. December, pp. 1–10, 2023, 10.3389/fpls.2023.1308528 Wu X, Fan X, Luo P, Das Choudhury S, Tjahjadi T, Hu C. From Laboratory to Field: Unsupervised Domain Adaptation for Plant Disease Recognition in the Wild. Plant Phenomics. 2023;5. 10.34133/plantphenomics.0038 . Meng Y, Zhan J, Li K, Yan F, Zhang L. A rapid and precise algorithm for maize leaf disease detection based on YOLO MSM. Sci Rep. 2025;15(1):6016. 10.1038/s41598-025-88399-1 . Shafik W, Tufail A, Namoun A, De Silva LC, A. RA, Apong HM. A Systematic Literature Review on Plant Disease Detection: Motivations, Classification Techniques, Datasets, Challenges, and Future Trends, IEEE Access , vol. 11, no. May, pp. 59174–59203, 2023, 10.1109/ACCESS.2023.3284760 Ngugi HN, Akinyelu AA, Ezugwu AE. Machine Learning and Deep Learning for Crop Disease Diagnosis, Agronomy , vol. 14, no. 12, 2024, [Online]. Available: https://www.mdpi.com/2073-4395/14/12/3001 Albert ETA, Bille NH, Leonard NME. Leveraging deep learning for plant disease identification: a bibliometric analysis in SCOPUS from 2018 to 2024. J Sci Agric. 2025;9:16–39. 10.25081/jsa.2025.v9.9412 . Hasan S, Jahan S, Islam MI. Disease detection of apple leaf with combination of color segmentation and modified DWT. J King Saud Univ - Comput Inf Sci. 2022;34(9):7212–24. https://doi.org/10.1016/j.jksuci.2022.07.004 . Abdou MA. Literature review: efficient deep neural networks techniques for medical image analysis, Neural Comput. Appl. , vol. 34, no. 8, pp. 5791–5812, 2022, 10.1007/s00521-022-06960-9 Roihan A, Sunarya PA, Rafika AS. Pemanfaatan Machine Learning dalam Berbagai Bidang: Review paper. IJCIT (Indonesian J Comput Inf Technol. 2020;5(1):75–82. 10.31294/ijcit.v5i1.7951 . Qotrunnada FM, Utomo PH. Metode Convolutional Neural Network untuk Klasifikasi Wajah Bermasker. Prisma. 2022;5:799–807. Arini A-, Wardhani LK, Octaviano D-. Perbandingan Seleksi Fitur Term Frequency & Tri-Gram Character Menggunakan Algoritma Naïve Bayes Classifier (Nbc) Pada Tweet Hashtag #2019gantipresiden. Kilat. 2020;9(1):103–14. 10.33322/kilat.v9i1.878 . Angga Irawan F, Sudarma M, Khrisne DC. Rancang Bangun Aplikasi Identifikasi Penyakit Tanaman Pepaya California Berbasis Android Menggunakan Metode CNN Model Arsitektur Squeezenet, Spektrum , vol. 8, no. 2, pp. 18–26, 2021, [Online]. Available: https://ojs.unud.ac.id/index.php/spektrum/article/view/75393 Setiawan W. Perbandingan Arsitektur Convolutional Neural Network Untuk Klasifikasi Fundus. J Simantec. 2020;7(2):48–53. 10.21107/simantec.v7i2.6551 . Angjaya HP, Gunadi K, Adipranata R. Pengenalan Penyakit pada Tanaman Pokok di Indonesia dengan Metode Convolutional Neural Network, J. Infra , 2021, [Online]. Available: http://publication.petra.ac.id/index.php/teknik-informatika/article/view/11426%0Ahttp://publication.petra.ac.id/index.php/teknik-informatika/article/viewFile/11426/10036 Vujović Ž. Classification Model Evaluation Metrics. Int J Adv Comput Sci Appl. 2021;12(6):599–606. 10.14569/IJACSA.2021.0120670 . Gao J, Ding M, Sun Q, Dong J, Wang H, Ma Z. Classification of Southern Corn Rust Severity Based on Leaf-Level Hyperspectral Data Collected under Solar Illumination. Remote Sens. 2022;14(11):2551. 10.3390/rs14112551 . Rozaqi AJ, Sunyoto A, Arief R. Deteksi Penyakit pada Daun Kentang Menggunakan Pengolahan Citra dengan Metode Convolutional Neural Network Detection of Potato Leaves Disease Using Image Processing with Convolutional Neural Network Methods. Creat Inf Technol J (CITEC JOURNAL). 2021;8(1):22–31. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7872919","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":562453175,"identity":"1b94572d-845c-4cde-91af-f23c0d7b90f3","order_by":0,"name":"Ismail Ismail","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIiWNgGAWjYBACNhCRUGDDwMAMZDwA8ZgZDIjQYpAG0ZJAjBYIMDgMoROgXLyK+fhPJ354YHA+cTs778EHCQx28gzszBvwO0wid7NEgsHtxJ3NfMkGCQzJhg3MbAUEtPBuAGvZcJjHTCKBgRmIeAh4n//s5h8JBudgWuqJ0MKQuw1oywGYlsNEaJHI3WaRYJBsvOEwyC8Gxw3bCPlFvv/s5ps/KuxkN5w/e/DBh4pqeX7+w/hDDAnwMIBjhI1Y9VAto2AUjIJRMAqwAAB/1zwG2ftOPAAAAABJRU5ErkJggg==","orcid":"","institution":"Universitas Lamappapoleonro","correspondingAuthor":true,"prefix":"","firstName":"Ismail","middleName":"","lastName":"Ismail","suffix":""},{"id":562453176,"identity":"710fec70-ab8f-4d21-bdff-0a55c718007a","order_by":1,"name":"Nursakti Nursakti","email":"","orcid":"","institution":"Universitas Lamappapoleonro","correspondingAuthor":false,"prefix":"","firstName":"Nursakti","middleName":"","lastName":"Nursakti","suffix":""},{"id":562453177,"identity":"09d96453-2bd4-4237-90ae-9cf2df625e83","order_by":2,"name":"Lut Faizal","email":"","orcid":"","institution":"Universitas Muhammadiyah Sinjai","correspondingAuthor":false,"prefix":"","firstName":"Lut","middleName":"","lastName":"Faizal","suffix":""},{"id":562453178,"identity":"cd2d17e3-686b-40b5-a49d-1fb267255924","order_by":3,"name":"Ibrahim Ibrahim","email":"","orcid":"","institution":"Universitas Lamappapoleonro","correspondingAuthor":false,"prefix":"","firstName":"Ibrahim","middleName":"","lastName":"Ibrahim","suffix":""}],"badges":[],"createdAt":"2025-10-16 03:38:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7872919/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7872919/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98768241,"identity":"200146ba-0905-4bc1-b330-3082caab3795","added_by":"auto","created_at":"2025-12-22 10:24:54","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1161533,"visible":true,"origin":"","legend":"","description":"","filename":"MenusciptRevisionV5.docx","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/75db47584899e4d359279aae.docx"},{"id":98768162,"identity":"e300f396-07a1-4b29-b1dc-f380a6e0a5d6","added_by":"auto","created_at":"2025-12-22 10:24:47","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7134,"visible":true,"origin":"","legend":"","description":"","filename":"6498ba54a654464d96891200fb2f29a1.json","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/32862b793058924730859676.json"},{"id":98768158,"identity":"b80bdc9f-7683-46fc-8b16-43d37d0c2d46","added_by":"auto","created_at":"2025-12-22 10:24:46","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":120627,"visible":true,"origin":"","legend":"","description":"","filename":"6498ba54a654464d96891200fb2f29a11enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/114a8d98b449e48b5a9ef352.xml"},{"id":98768152,"identity":"074a11c9-8a23-4151-849c-7035914d840e","added_by":"auto","created_at":"2025-12-22 10:24:40","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69299,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/6db01f74517470ddd6be25a3.png"},{"id":98768155,"identity":"a9248c18-2361-473c-9d09-f47d0d96b90b","added_by":"auto","created_at":"2025-12-22 10:24:41","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":156139,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/e1bfdad696a3e19babeb26b0.png"},{"id":98768154,"identity":"e36d99de-4d94-476e-a1e4-baad15ea4ca9","added_by":"auto","created_at":"2025-12-22 10:24:41","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44508,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/aca45a0cd6e33941ca7da0dd.png"},{"id":98768153,"identity":"56343313-52e9-460b-8bbc-e2513af2d5ee","added_by":"auto","created_at":"2025-12-22 10:24:41","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63808,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/72df4a05922b7c992ed95ec0.png"},{"id":98780191,"identity":"9f923539-b3c0-4b4f-9ea3-97458a4e2ccb","added_by":"auto","created_at":"2025-12-22 12:31:09","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":370241,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/ac2ee2e73f15c52846e6f5fd.png"},{"id":98768150,"identity":"a903de1e-dc2f-44ef-ac2d-615905bf8b36","added_by":"auto","created_at":"2025-12-22 10:24:40","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":33759,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/90600e6fd6c2e3d3e2725ff6.png"},{"id":98768156,"identity":"108fed36-3f66-4ecc-80d9-f1fb26244e7a","added_by":"auto","created_at":"2025-12-22 10:24:41","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40644,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/f1e08d57fbcd44195af0a7d1.png"},{"id":98768145,"identity":"26af826c-0539-4ce1-b81e-3a6649b4ba85","added_by":"auto","created_at":"2025-12-22 10:24:39","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24391,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/0257f9cdaafc066f2fc0d288.png"},{"id":98768147,"identity":"18020c08-b181-47be-9c42-05ffce78da88","added_by":"auto","created_at":"2025-12-22 10:24:39","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30686,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/4b290257bb23c35b055c20fa.png"},{"id":98768160,"identity":"d3e1a1f3-2645-4d94-8de0-a6ad5f709dd9","added_by":"auto","created_at":"2025-12-22 10:24:47","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":119107,"visible":true,"origin":"","legend":"","description":"","filename":"6498ba54a654464d96891200fb2f29a11structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/5a1db5c69b578b3c80b6ac76.xml"},{"id":98780432,"identity":"7eb42df9-95f1-43c4-8e1d-5b01a51e8a11","added_by":"auto","created_at":"2025-12-22 12:31:20","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":132668,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/87a2889ff6a41d1c335d567a.html"},{"id":98768146,"identity":"9f522176-ea47-46fe-802e-e546cba3a681","added_by":"auto","created_at":"2025-12-22 10:24:39","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":271791,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eFlow diagram of Research Method\u003c/em\u003e [28]\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/b269eb037096968f3d193840.jpeg"},{"id":98768243,"identity":"26986cd0-6b23-4ef4-b185-db9b3a2bfe94","added_by":"auto","created_at":"2025-12-22 10:24:54","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":433556,"visible":true,"origin":"","legend":"\u003cp\u003e(a) \u003cem\u003eBlight,\u003c/em\u003e (b) \u003cem\u003eCommon Rust,\u003c/em\u003e (c) \u003cem\u003eGray Leaf Spot,\u003c/em\u003e (d) \u003cem\u003eHealthy \u003c/em\u003e[30]\u003c/p\u003e\n\u003cp\u003eSource: \u003ca href=\"https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\"\u003ehttps://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\u003c/a\u003e\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/0a77b65776c27bc4f5a615a8.jpeg"},{"id":98768148,"identity":"e9503348-0e36-4450-93d2-5451e80e75eb","added_by":"auto","created_at":"2025-12-22 10:24:39","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":160712,"visible":true,"origin":"","legend":"\u003cp\u003eLayer on Convolutional Neural Network[34]\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/d84f332ff7e95387e93e608c.jpeg"},{"id":98768245,"identity":"fee811ae-daa0-477b-bd93-5f525db6fa99","added_by":"auto","created_at":"2025-12-22 10:24:55","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":223953,"visible":true,"origin":"","legend":"\u003cp\u003eProposed architecture [37]\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/546c1f0ea2200513c325373f.jpeg"},{"id":98768163,"identity":"4f0a83b5-5b59-4317-8782-6c101bc43277","added_by":"auto","created_at":"2025-12-22 10:24:47","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":451377,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eTraining Dataset\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/ccffbf07ae4fc9d049da9b68.jpeg"},{"id":98768197,"identity":"72d84e0f-9408-4738-a94f-7ff134d9bfa4","added_by":"auto","created_at":"2025-12-22 10:24:48","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":138547,"visible":true,"origin":"","legend":"\u003cp\u003eTraining and Validation Accuracy of CNN Model Over 100 Epochs\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/f5ffcd19faf903b057d7e015.jpeg"},{"id":98768238,"identity":"b756fa03-1076-48ae-805a-fbe07d6bff43","added_by":"auto","created_at":"2025-12-22 10:24:53","extension":"jpeg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":186026,"visible":true,"origin":"","legend":"\u003cp\u003eModel Loss Over 100 Training Epochs\u003c/p\u003e","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/c7037b7809382c2d9ebc5861.jpeg"},{"id":98768196,"identity":"61167f42-0170-4b1a-866d-11dfe8d35c46","added_by":"auto","created_at":"2025-12-22 10:24:48","extension":"jpeg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":92579,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion Matrix of CNN Model for Corn Leaf Disease Classification\u003c/p\u003e","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/98e05a9dadda75ab3650bab4.jpeg"},{"id":98768242,"identity":"a7175370-ab97-47b8-8808-88308ac2e15a","added_by":"auto","created_at":"2025-12-22 10:24:54","extension":"jpeg","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":131629,"visible":true,"origin":"","legend":"\u003cp\u003eClassification Report Metrics for Corn Leaf Disease Detection\u003c/p\u003e","description":"","filename":"floatimage9.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/72db937dec178b2b96b21480.jpeg"},{"id":102906099,"identity":"8a9dd3c0-c852-4db6-ade8-cdc61b00d075","added_by":"auto","created_at":"2026-02-18 09:12:28","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2737481,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7872919/v1/68725116-a5c4-41d3-8615-94bc534664d6.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Deep Learning Based Multiclass Detection of Corn Leaf Diseases Using a Convolutional Neural Network","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eCorn is a food plant that occupies the third position in the world after rice and wheat, and ranks second in Indonesia after rice [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Corn contains carbohydrates, which have the potential to be developed as functional food products because corn contains essential amino acids, minerals, dietary fiber, essential fatty acids, and others needed by the body. Increasing consumer demand, of course, will determine the production quantity for corn cultivation [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. According to research, there are various aspects that affect production in corn cultivation, namely seeds, land area, fertilization accuracy, and labor. Fertilization aims to improve crop quality and prevent pests or corn diseases. Pests can cause crop yields to be less than optimal, and even harvest failure can occur, so maximizing the amount of corn production requires the right steps or cultivation to anticipate pest attacks that can cause various diseases in corn plants [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Corn plant diseases can be identified manually with the naked eye, because the color of the corn leaves will change each time it is affected by the disease. However, the color of the leaves is rather difficult to distinguish because the color is almost the same, apart from that, of course, it will take a long time if the plants are identified in large quantities, especially since each individual has a different color assessment [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIdentifying this disease requires an expert, because the disease's lethal characteristics and lack of similarities can give the wrong impression. One method that can be used to classify leaf diseases is using image datasets, namely deep learning[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Deep Learning is a branch of broader machine learning using artificial neural networks. Deep learning architectures can be artificial neural networks, recurrent neural networks, and convolutional neural networks that have been applied to areas such as pattern recognition, speech recognition, natural language processing, audio recognition, social network filtering, machine translation, bioinformatics, medical image analysis, and other fields with satisfactory results [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. There has been a lot of deep learning development, such as CNN, which has been widely researched in terms of making models to get high accuracy values.\u003c/p\u003e \u003cp\u003eThis CNN is a refinement of the multi-layer perceptron or MLP algorithm, built to manage 2-dimensional data. CNN belongs to the deep neural network category because it is a fairly good network and has been widely applied for processing image data. In the CNN Algorithm, there is an arrangement of text into a matrix, with each row representing a word embedding, a word, or a character [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. By utilizing the Convolutional Neural Network (CNN) method as one of the Machine Learning technologies, the problem of classifying the image of corn plant leaves will be easier to solve [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Research has been carried out using artificial neural network methods based on backpropagation algorithms, with an accuracy rate of 90% and an error of 10% [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The next researcher detected diseases in corn plants based on image processing using the color moments and GLCM methods. This method achieves an accuracy rate of 89.375% [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Another study with a different object, namely the detection of disease in potatoes using the CNN method, resulted in a training accuracy of 95% with a validation accuracy of 94% [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWhile prior studies have applied CNN and other machine learning models for plant disease detection, most have focused on binary classification or limited crop types. Moreover, many works lack validation using robust performance metrics or rely on manually engineered features [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. This study presents a comprehensive multiclass classification of corn leaf diseases using an end-to-end CNN model trained from scratch on a curated public dataset [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. The novelty of this work lies in: 1) Application of a custom CNN architecture specifically tailored for corn leaf imagery rather than relying on generic or pre-trained models. 2) Multiclass classification across four disease categories, including subtle differentiations between Blight, Rust, and Gray Leaf Spot, which are visually similar. 3) Quantitative evaluation using a full suite of metrics (precision, recall, F1-score, confusion matrix) to ensure class-level performance is well understood. 4) Implementation of a scalable preprocessing pipeline, enabling reproducibility and real-world deployment in smart agriculture systems. This research contributes to the development of automated, accurate, and scalable solutions for disease detection in crops, particularly in contexts with limited access to expert agronomists\u003c/p\u003e"},{"header":"2. Related work","content":"\u003cp\u003eSeveral studies have explored the application of machine learning and deep learning techniques to plant disease classification using leaf imagery. Traditional image processing approaches, such as those based on color moment and texture features, have demonstrated moderate success[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. For example, employed the Color Moment and Gray-Level Co-occurrence Matrix (GLCM) methods to detect rubber plant diseases, achieving an accuracy of 89.37%. However, these approaches often rely on handcrafted features, which may not generalize well across different plant species and lighting conditions. With the emergence of deep learning, Convolutional Neural Networks (CNNs) have become a dominant approach for image-based classification tasks[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Applied a pre-trained deep learning model to classify corn leaf diseases and achieved an accuracy of 90%, indicating the potential of CNNs for agricultural applications. Similarly, developed a CNN model for detecting diseases in potato leaves, reporting a training accuracy of 95% and a validation accuracy of 94% [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. While promising, these studies were either limited to binary classification or involved crops other than corn, which poses a challenge for generalization in corn-specific pathology [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eExplored the VGG16 pre-trained CNN architecture for grape leaf disease classification, which demonstrated high accuracy. However, pre-trained networks such as VGG16, while effective, often contain large parameter sizes and are not optimized for lightweight deployment in agricultural field conditions. Additionally, their reliance on transfer learning may introduce bias when applied to non-natural image datasets such as plant leaves. More recent efforts have aimed at developing mobile-based or lightweight CNN variants tailored for specific plants [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Introduced a CNN-Squeeze Net-based application for detecting papaya diseases, showing that custom architectures can outperform generic models in domain-specific tasks. Despite this progress, most existing models are either trained on small, imbalanced datasets or lack rigorous evaluation through per-class metrics such as precision, recall, and F1-score. In the context of corn plant disease detection, a significant gap remains in the development of a CNN model specifically trained and evaluated on a multiclass corn leaf disease dataset. Moreover, the use of balanced, high-resolution image data from open-source repositories like Kaggle has not been fully leveraged to build robust models tailored to the characteristics of corn leaves, which often display overlapping visual symptoms such as rust, blight, and gray leaf spot [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eRecent studies increasingly show that Vision Transformers (ViT) and hybrid CNN-Transformer architectures can outperform or at least match classic CNN backbones for crop disease recognition\u0026mdash;especially on complex, in-field images. Hybrid pipelines that fuse local CNN features with ViT\u0026rsquo;s global context have improved transferability and feature discrimination under variable illumination and background clutter, narrowing the lab-to-field gap (e.g., MViT\u0026thinsp;+\u0026thinsp;CNN frameworks)[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. Comprehensive 2025 surveys of maize leaf disease detection echo this trend, reporting stronger benchmarks from Swin/ViT variants and CNN-ViT hybrids compared with pure CNNs, while also highlighting dataset curation and augmentation as performance drivers[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Recent transformer-centric solutions (e.g., Swin-based or ViT-enhanced classifiers) continue to set competitive results on PlantVillage-style corpora and mixed field datasets, underscoring the growing role of attention mechanisms in modeling subtle visual symptoms on maize leaves [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn parallel, there\u0026rsquo;s a clear push toward lightweight models suitable for mobile and edge deployment so that diagnosis can run on-device without cloud connectivity. 2025 work proposes compact MobileNet-style variants (e.g., RTR_Lite_MobileNet) that retain accuracy while cutting parameters and latency, making them attractive for farm-side use in low-resource settings[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Similar efforts in maize-specific classifiers emphasize interpretability and efficiency to aid extension workers and farmers, balancing accuracy with model size and explainability. More broadly, edge-computing studies in plant phenotyping argue that optimizing architectures and memory footprints is essential for practical, field-ready systems\u0026mdash;reinforcing the relevance of lightweight CNNs and quantization/pruning for deployment [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFinally, to improve robustness and generalization beyond curated datasets, recent work explores domain adaptation, few-shot learning, and real-time detection. Unsupervised domain adaptation methods mitigate performance drops when moving from lab images to diverse field conditions with shifting backgrounds and lighting[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Cross-domain few-shot frameworks enable fast adaptation to unseen farms or newly emerging maize diseases with very limited labeled samples, which is crucial where expert annotation is scarce. Alongside classification, object-detection pipelines tailored to maize leaves (e.g., YOLO-style models with multi-scale kernels) enable real-time localization of symptomatic regions directly in the field, supporting precision scouting and timely intervention [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study addresses the aforementioned gaps by proposing a custom CNN model trained from scratch on a curated and balanced dataset of corn leaf images. It employs a standardized preprocessing pipeline and conducts a comprehensive evaluation using class-wise precision, recall, and F1-score. Unlike prior works, the proposed approach focuses specifically on \u003cb\u003emulticlass disease classification in corn\u003c/b\u003e, which is crucial for precision agriculture applications in low-resource settings.\u003c/p\u003e"},{"header":"3. Research Method","content":"\u003cp\u003eThis research had several stages in perfecting the grouping or classification stage of corn leaf disease, including problem identification, literature study, dataset collection, dataset processing, CNN, and results evaluation. Solving the problem using the Convolutional Neural Network (CNN) method in identifying corn plant leaves [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe stages of the research methodology were: 1) Problem Identification. At this stage, data collection and problem analysis were carried out. 2) Literature study, After collecting the dataset used in data processing, data processing was carried out using the Convolution Neural Network algorithm. 3) Evaluation of Results, After processing the data using the Convolution Neural Network algorithm, the results were then evaluated. [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The following was the methodology in this study, which is described in the form of a diagram flow as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCollection of Datasets\u003c/p\u003e \u003cp\u003eThe data used in this study was the corn leaf dataset obtained from the website with the URL: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\u003c/span\u003e\u003cspan address=\"https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e accessed on Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, 11 February 2025.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe corn leaf datasets [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eClass\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eBlight\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.146\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eCommon Rust\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.306\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eGray Leaf Spot\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e574\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eHealthy\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.162\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cem\u003eTotal\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.188\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTypes of leaf diseases in corn plants include blight, common rust, grey leaf spot, and healthy. Blight-type corn plant disease is an attack that occurs early in growth, causing the plant to wither and die. Common rust is a leaf rust disease on corn caused by a fungus. Gray leaf spot is an infectious disease that can be carried by wind or rain splashes and can cause the first infection in corn plants. Healthy is a disease characterized by a change in leaf color from pale green to yellow, accompanied by regular linear lines or parallel wavy edges. The following is Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, of a disease on corn leaves [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSource: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\u003c/span\u003e\u003cspan address=\"https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003cp\u003eDataset Processing\u003c/p\u003e \u003cp\u003eData preprocessing serves as a crucial initial step in preparing image datasets for classification tasks, particularly in deep learning applications. In this study, the preprocessing phase involved partitioning the dataset into distinct subsets for training and testing, following an 80:20 ratio to ensure an adequate representation of all classes during model learning and evaluation. Additionally, to maintain consistency and optimize computational efficiency, all images were resized to a uniform resolution of 150 \u0026times; 150 pixels. This resizing not only standardizes the input dimensions across the dataset but also ensures compatibility with the input layer requirements of the Convolutional Neural Network (CNN) model. Normalization techniques were also applied to scale pixel values within a range suitable for neural network training, thereby improving convergence during optimization [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. These preprocessing steps are essential for minimizing data inconsistencies, reducing noise, and enhancing the model\u0026rsquo;s ability to learn discriminative features effectively.\u003c/p\u003e \u003cp\u003e \u003cem\u003eConvolutional Neural Network (CNN)\u003c/em\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eA convolutional neural network is a type of neural network that can be used in image data. CNN is also commonly used to find and identify objects in the form of images. CNN is used in solving the problem of mapping corn leaf area, so that the data can be used to solve problems in classifying corn diseases [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. CNN is used to group images and is also used to detect objects in images [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. CNN has a series of 3D neurons (width, height, and intensity) which are the shape of the layers, while the intensity or depth is the number of layers. There are generally 2 (two) layers in CNN, namely: feature learning and classification, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe layers in the Convolutional Neural Network are as follows:\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003e1. \u003cem\u003eLayer feature learning\u003c/em\u003e/image feature extraction layer, composed of several layers, and each layer is composed of various neurons that are connected to the local location of the previous layer. The first type of layer is the convolution layer, and the second layer is the pooling layer, which is alternately located.\u003c/p\u003e \u003cp\u003e2. Layer classification/classification, systematically from various layers, with each layer being composed of neurons that are fully connected to other layers. The output of this layer is in the form of class scoring for grouping [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe research object was carried out to identify disease problems in the leaves of corn plants using the CNN algorithm with its architecture [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. A convolutional neural network architecture is commonly used in image data. CNN is not much different from neural networks in general. CNN also consists of neurons that have weights, biases, and activation functions. The proposed CNN architecture for identifying corn leaf diseases is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]:\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eConfusion Matrix\u003c/em\u003e \u003c/p\u003e \u003cp\u003eThe form of the method for evaluating classification is commonly called the confusion matrix and is usually used for calculating recall, error rate, precision, and accuracy. Recall is the ability of the system to determine relevant items, which is defined as the percentage of documents that apply to the query. Accuracy is a comparison where the overall number of errors is correctly identified, and the things that are identified incorrectly are divided by the total number of all cases. Precision evaluates whether the system is capable of detecting the most relevant rating [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. The confusion matrix is ​​a table that states the classification of the correct number of test data and the incorrect number of test data [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. The following Table \u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e is a multiclass confusion matrix.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cem\u003eMulticlass Confusion Matrix\u003c/em\u003e (MCM)[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c3\"\u003e \u003cp\u003ePREDICTION\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003ePOSITIF\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003eNEGATIF\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eNEUTRAL\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003e\u003cb\u003eACTUAL\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003ePOSITIF\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTPos\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFPosNeg\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFPosNet\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eNEGATIF\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFNegPos\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTNeg\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFNegNet\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eNEUTRAL\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFNetpos\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFNetNeg\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTNet\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eMeanwhile, to calculate the accuracy evaluated in terms of overall accuracy (OA), precision, recall, and F1-score can be calculated using the equations (1), (2), (3), and (4) [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]:\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Tabc\" border=\"1\"\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOA = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{\\text{T}\\text{P}\\:+\\:\\text{T}\\text{N}}{\\text{T}\\text{P}\\:+\\:\\text{F}\\text{P}\\:+\\:\\text{T}\\text{N}\\:+\\:\\text{F}\\text{N}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrecision = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{\\text{T}\\text{P}}{\\text{T}\\text{P}\\:+\\:\\text{F}\\text{P}\\:}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRecall = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{\\text{T}\\text{P}}{\\text{T}\\text{P}\\:+\\:\\text{F}\\text{N}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eF1 - Score = \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\frac{2\\:\\times\\:\\:\\text{T}\\text{P}}{2\\:\\times\\:\\:\\text{T}\\text{P}\\:+\\:\\text{F}\\text{P}\\:+\\:\\text{F}\\text{N}}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eHere is the Equation Information:\u003c/p\u003e \u003cp\u003eOA\u0026thinsp;=\u0026thinsp;\u003cem\u003eOverall accuracy\u003c/em\u003e\u003c/p\u003e \u003cp\u003eTP\u0026thinsp;=\u0026thinsp;\u003cem\u003eTrue Positive\u003c/em\u003e\u003c/p\u003e \u003cp\u003eTN\u0026thinsp;=\u0026thinsp;\u003cem\u003eTrue Negative\u003c/em\u003e\u003c/p\u003e \u003cp\u003eFN\u0026thinsp;=\u0026thinsp;\u003cem\u003eFalse Negative\u003c/em\u003e\u003c/p\u003e \u003cp\u003eFP\u0026thinsp;=\u0026thinsp;\u003cem\u003eFalse Positive\u003c/em\u003e\u003c/p\u003e"},{"header":"4. Results","content":"\u003cp\u003eThe classification model was trained and evaluated using a dataset of 4,188 corn leaf images, which were divided into training and testing subsets using an 80:20 split. The training set comprised 3,348 images, while the remaining 840 images were used for testing. The proposed Convolutional Neural Network (CNN) architecture was trained for 100 epochs using a batch size of 32, and validation was conducted using a batch size of 8. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows the output of the designed fit model.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe training log presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e illustrates the model's performance over 100 epochs, demonstrating a consistent trend of convergence and generalization. The final training accuracy reached \u003cb\u003e95.34%\u003c/b\u003e, while the corresponding validation accuracy stabilized at \u003cb\u003e93.93%\u003c/b\u003e, indicating that the model achieved high predictive performance with minimal overfitting. Notably, the validation accuracy remained above 92% throughout the final epochs, suggesting robust learning across all classes. Although a slight increase in validation loss was observed near the final iterations\u0026mdash;rising to \u003cb\u003e0.3617\u003c/b\u003e\u0026mdash;the model maintained stable validation accuracy, implying resilience against overfitting. The progressive reduction in training loss, culminating at \u003cb\u003e0.1219\u003c/b\u003e, further confirms the model's ability to learn discriminative features effectively. These results substantiate the model\u0026rsquo;s capability to generalize well to unseen data, thereby supporting its applicability for real-world classification of corn leaf diseases.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e illustrates the model accuracy progression over 100 training epochs for both training and validation sets. The training accuracy demonstrates a steady upward trend, ultimately reaching approximately \u003cb\u003e95%\u003c/b\u003e, while the validation accuracy closely follows, stabilizing around \u003cb\u003e93\u0026ndash;94%\u003c/b\u003e. The early epochs display notable fluctuations in validation performance, which is typical as the model adjusts its internal parameters; however, the fluctuations gradually decrease, indicating improved stability and generalization. The convergence of both curves after epoch 30, with minimal divergence between training and validation accuracy, suggests that the model effectively learned the data distribution without overfitting. This level of consistency confirms the robustness of the proposed CNN architecture and highlights its potential for accurate classification of corn leaf diseases in real-world applications.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e presents the training and validation loss curves over 100 epochs, offering insights into the model\u0026rsquo;s learning dynamics and generalization capability. The training loss exhibits a smooth and steady decline, reaching a final value below 0.15, which indicates that the model effectively minimized error on the training set. In contrast, the validation loss initially follows a similar downward trend but begins to fluctuate after approximately 20 epochs, stabilizing around 0.35. These fluctuations are typical in deep learning models trained on natural image data and may reflect minor variations in batch composition or limited noise robustness. Despite the variability in validation loss, no significant divergence from the training loss is observed, suggesting that overfitting remains well-controlled. The consistently low training loss and moderately stable validation loss confirm that the model achieved a balanced fit and retained its ability to generalize across unseen data, thus validating its reliability for multiclass corn disease classification tasks.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClassification Report [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c6\" namest=\"c3\"\u003e \u003cp\u003ePREDICTION\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSupport\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eBlight\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eCommon Rust\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eGray Leaf Spot\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003eHealthy\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eACTUAL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBlight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e211\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e230\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCommon Rust\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e252\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e262\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGray Leaf Spot\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e115\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHealthy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e231\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e233\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e235\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e266\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e108\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e231\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e840\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAccuracy = (211\u0026thinsp;+\u0026thinsp;252\u0026thinsp;+\u0026thinsp;95\u0026thinsp;+\u0026thinsp;231) / 840\u0026thinsp;=\u0026thinsp;0,939\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eRecal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eF1-Score\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eBlight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e211/235\u0026thinsp;=\u0026thinsp;0,897\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e211/230\u0026thinsp;=\u0026thinsp;0,917\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e(2\u0026times;0,897\u0026times;0,917)/ (0,897\u0026thinsp;+\u0026thinsp;0,917)\u0026thinsp;=\u0026thinsp;0,906\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eCommon Rust\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e252/266\u0026thinsp;=\u0026thinsp;0,947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e252/262\u0026thinsp;=\u0026thinsp;0,961\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e(2\u0026times;0,947\u0026times;0,961)/ (0,947\u0026thinsp;+\u0026thinsp;0,961)\u0026thinsp;=\u0026thinsp;0,953\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eGray Leaf Spot\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e95/108\u0026thinsp;=\u0026thinsp;0,879\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e95/115\u0026thinsp;=\u0026thinsp;0,826\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e(2\u0026times;0,879\u0026times;0,826)/ (0,879\u0026thinsp;+\u0026thinsp;0,826)\u0026thinsp;=\u0026thinsp;0,851\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eHealthy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e231/231\u0026thinsp;=\u0026thinsp;1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e231/233\u0026thinsp;=\u0026thinsp;0,991\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e(2\u0026times;1 \u0026times;0,991)/ (1\u0026thinsp;+\u0026thinsp;0,991)\u0026thinsp;=\u0026thinsp;0,995\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the detailed classification report, highlighting the model\u0026rsquo;s predictive performance across four categories of corn leaf health: Blight, Common Rust, Gray Leaf Spot, and Healthy. The overall accuracy of the model reached 93.9%, demonstrating strong generalization capability. Notably, the Healthy class achieved a perfect precision score of 1.00, with a recall of 0.991, resulting in an exceptional F1-score of 0.995, indicating that the model identified healthy leaves with near-perfect reliability. The Common Rust class also performed robustly, with a precision of 0.947, a recall of 0.961, and an F1-score of 0.953, suggesting the model's effectiveness in recognizing this disease despite its visual similarity to other conditions. Although Gray Leaf Spot exhibited slightly lower metrics (F1-score: 0.851), the results remain acceptable given the class\u0026rsquo;s lower sample size and inter-class feature overlap.\u003c/p\u003e \u003cp\u003eBlight showed balanced performance with a precision of 0.897 and a recall of 0.917. These findings indicate that the model performs well across all classes, particularly in high-confidence identification of healthy and rust-infected leaves, making it a reliable tool for automated disease diagnosis in corn cultivation. This study uses the sklearn Python module to visualize the Confusion Matrix so that it can provide information in the form of system detection results for each class tested based on image data testing. The results of system detection can be seen in the following figure.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e is the result of a visualization of the 4\u0026times;4 confusion matrix table, which displays the prediction results of the system. The explanation for each class is as follows:\u003c/p\u003e \u003cp\u003e1. Blight class: From a total of 230 test data, the system recognizes 211 images according to their class, 9 data points that should be classified as blight class but are detected as common rust class, and 10 data points are detected as grey leaf spot class.\u003c/p\u003e \u003cp\u003e2. \u003cem\u003eCommon rust\u003c/em\u003e class: From a total of 262 test data, the system recognizes 252 images according to their class, 7 data points that should be of the common rust blight class but are detected as the blight class, and 3 data points are detected as the grey leaf spot class.\u003c/p\u003e \u003cp\u003e3. \u003cem\u003eGray leaf spot\u003c/em\u003e class: From a total of 115 test data, the system recognizes 95 images according to their class, 17 data points that should be of the grey leaf spot class but are detected as blight class, and 3 data points are detected as common rust class.\u003c/p\u003e \u003cp\u003e4. \u003cem\u003eHealthy rust\u003c/em\u003e class: From a total of 233 test data, the system recognizes 231 images according to their class, and 2 data points that do not match their class because they are detected as common rust class\u003c/p\u003e \u003cp\u003eIn addition to visualizing the test results in the form of a confusion matrix table, the system is also tested for performance by calculating the accuracy, precision, recall, and F1-score values. The performance test results can be seen in the following Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e:\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAccuracy, precision, recall, and F1-score values ​​are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e as a result of testing system performance using the scikit learn library. In testing 230 blight class images, a precision value of 0.90 was obtained, a recall value of 0.92, and an F1-score of 0.91. The precision value for the common rust class is 0.95, the recall value is 0.96, and the F1-score is 0.95. The precision value for the gray leaf spot class is 0.88, the recall value is 0.83, and the F1-score is 0.85. In the healthy class, the precision value is 1, the recall is 0.99, and the F1-score is 1. The average value of the accuracy of the system for detecting corn leaf disease is 94%\u003c/p\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThis research was designed to address the need for accurate, efficient, and automated detection of corn leaf diseases, particularly in the context of smallholder farming in Soppeng Regency. The primary objective was to develop and evaluate a Convolutional Neural Network (CNN) model capable of classifying multiple corn leaf diseases\u0026mdash;such as rust, blight, and gray leaf spot\u0026mdash;based on image data. By leveraging high-resolution, balanced datasets from open-source repositories, the research aimed to overcome limitations found in previous studies, such as reliance on binary classification, small or imbalanced datasets, and lack of field-ready deployment strategies.\u003c/p\u003e \u003cp\u003eThe experimental results demonstrate that the proposed CNN model achieved high classification accuracy across all targeted corn leaf disease categories, with robust performance in both training and validation phases. The model consistently maintained strong per-class metrics, including precision, recall, and F1-score, indicating its reliability in distinguishing between diseases with visually overlapping symptoms. This outcome confirms that a properly optimized CNN architecture, trained on a well-curated dataset, can effectively handle the complexities of corn pathology in diverse field conditions. The novelty of this research lies in its integration of a CNN architecture specifically trained for multiclass corn disease detection using a balanced and high-quality dataset. Unlike many earlier works that focused on single-disease detection or crops other than corn, this study targets multiple diseases simultaneously while mitigating class imbalance through data preprocessing techniques. Furthermore, the approach emphasizes high-resolution image analysis, enabling finer feature extraction and improved discrimination between diseases with similar visual characteristics.\u003c/p\u003e \u003cp\u003eThis research contributes to the agricultural technology domain by providing an end-to-end framework for automated disease detection that is adaptable to practical farming environments. The proposed CNN model can be integrated into mobile or edge computing platforms, enabling farmers and agricultural extension workers to diagnose plant diseases rapidly in the field. In addition, the study offers a reproducible methodology for dataset preparation, model training, and evaluation, which can serve as a reference for future research in smart agriculture and precision farming. From a sustainability perspective, the system developed in this study has the potential to reduce dependency on expert manual inspections, thereby lowering operational costs and enabling faster disease management interventions. By facilitating early and accurate detection, it can help prevent severe crop losses, reduce pesticide misuse, and promote more targeted and environmentally friendly agricultural practices. The adaptability of the framework to other crops and regions ensures that the impact of this research can extend beyond corn cultivation in Soppeng Regency, contributing to broader goals in sustainable food production.\u003c/p\u003e \u003cp\u003eThe proposed CNN-based system demonstrates a reliable, accurate, and efficient solution for detecting multiple corn leaf diseases. The results validate the research objectives and highlight the model\u0026rsquo;s potential for real-world agricultural deployment. The novelty in dataset utilization, the methodological rigor, and the practical adaptability make this work a valuable contribution to both academic research and applied farming practices. Future work could focus on integrating real-time image acquisition, expanding the dataset with locally sourced images, and optimizing the model for even greater efficiency in low-resource settings.\u003c/p\u003e \u003cp\u003eBased on the research findings, the proposed Convolutional Neural Network (CNN) model has proven effective in accurately classifying multiple corn leaf diseases, including rust, blight, and gray leaf spot, using high-resolution and balanced image data. The model achieved high accuracy and strong per-class performance metrics, demonstrating its reliability in distinguishing diseases with similar visual characteristics. These results confirm that a well-trained CNN architecture, supported by quality datasets, can overcome the limitations of traditional methods and earlier studies. Furthermore, the model\u0026rsquo;s potential for integration into mobile or edge-based platforms makes it a practical tool for real-time disease detection in the field, supporting timely decision-making and contributing to sustainable agricultural practices.\u003c/p\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis research succeeded in implementing machine learning using the Convolutional Neural Network (CNN) algorithm with the sklearn Python library to produce a good level of training accuracy. Presents a high-performing Convolutional Neural Network (CNN) model designed specifically for the multiclass classification of corn leaf diseases using image data. The model was trained and validated on a curated dataset of 4,188 corn leaf images, comprising four classes: Blight, Common Rust, Gray Leaf Spot, and Healthy. The final model achieved an overall accuracy of 93.9%, with class-specific F1-scores ranging from 0.851 to 0.995. The Healthy and Common Rust classes exhibited the highest performance, indicating the model\u0026rsquo;s strong ability to identify clear visual symptoms. Despite some challenges in distinguishing Gray Leaf Spot due to its visual similarity to other classes, the results demonstrate the model\u0026rsquo;s robustness and generalizability. The use of a custom-built CNN architecture, combined with a systematic preprocessing pipeline, significantly enhanced learning efficiency and prediction consistency. These findings validate the feasibility of applying deep learning for automated, precise, and scalable disease diagnosis in corn cultivation, offering valuable potential for integration into smart agriculture systems, especially in low-resource settings.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflict of Interest\u003c/h2\u003e \u003cp\u003eThe authors hereby declare that there are no known financial, personal, or professional conflicts of interest that could have influenced the conduct, analysis, or reporting of this research. All stages of the study, including data collection, model development, interpretation of results, and manuscript preparation, were carried out independently and without any undue influence from external parties. The funding provided by Lamappapoleonro University was granted to support academic research and did not affect the objectivity, integrity, or transparency of the work presented in this article. As such, the authors affirm that the findings and conclusions are solely the result of scholarly investigation.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eEthics approval\u003c/strong\u003e \u003cp\u003eEthics statement Not applicable because this research did not involve human participants, animals, or personally identifiable data. All datasets used were obtained from open access repositories and are freely available for academic and research purposes. This study obtained consent from all participants prior to data collection, and confidentiality was maintained throughout the research.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent to participate\u003c/strong\u003e \u003cp\u003eAll authors have read and approved the final version of this manuscript and consent to its submission for publication. Each author confirms their active participation in the research, data analysis, and manuscript preparation in accordance with ethical publication standards. The authors collectively agree to be accountable for all aspects of the work, ensuring the integrity and accuracy of the published article.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent to publish\u003c/strong\u003e \u003cp\u003eThis research did not involve human participants, animals, or personally identifiable data. Therefore, consent to publish the research results presented in this article is not required.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eClinical Trial Number\u003c/h2\u003e \u003cp\u003eThis study did not involve any clinical trials; therefore, a clinical trial registration number is not applicable. The research was conducted using non-clinical methods and did not include any interventions or procedures involving human participants or patients. All data were collected and analyzed in accordance with ethical research standards without the need for clinical trial registration.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAll authors contributed significantly to the development of this research. **Ismail** formulated the study concept, designed the CNN model architecture, and led the data analysis and interpretation. He also prepared the initial draft of the manuscript. **Nursakti** was involved in the preprocessing of the dataset. **L Faizal** performed the performance evaluation and refinement of the deep learning framework. **Ibrahim** provided critical revisions to the manuscript and contributed to the discussion and conclusion sections. All authors reviewed and approved the final version of the manuscript and agreed to be accountable for all aspects of the work to ensure its accuracy and integrity.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors would like to express their sincere gratitude to the Rector of Lamappapoleonro University for the invaluable support and funding provided for this research. The opportunity and resources extended by the university leadership have played a pivotal role in the successful completion of this study. This support reflects the institution's strong commitment to advancing research and innovation in the field of agricultural technology and artificial intelligence\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data used in this study comes from a publicly available and curated image repository. The dataset was obtained from the \u0026ldquo;Corn or Maize Leaf Disease Dataset\u0026rdquo; . The data source can be seen in the Kaggle \u0026ndash; https://www.kaggle.com/smaranjitghose/corn-or-maize-leaf-disease-dataset\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRozi F, et al. Indonesian market demand patterns for food commodity sources of carbohydrates in facing the global food crisis. Heliyon. 2023;9(6):e16809. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.heliyon.2023.e16809\u003c/span\u003e\u003cspan address=\"10.1016/j.heliyon.2023.e16809\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu B-G, Chen X-X, Zhou C-X, Ding T-B, Wang Z-H, Zou C-Q. Nutritional composition of maize grain associated with phosphorus and zinc fertilization. J Food Compos Anal. 2022;114:104775. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jfca.2022.104775\u003c/span\u003e\u003cspan address=\"10.1016/j.jfca.2022.104775\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorales H, Perfecto I, Ferguson B. Traditional fertilization and its effect on corn insect populations in the Guatemalan highlands. Agric Ecosyst Environ. 2001;84(2):145\u0026ndash;55. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S0167-8809(00)00200-0\u003c/span\u003e\u003cspan address=\"10.1016/S0167-8809(00)00200-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhan SU, Alsuhaibani A, Alabduljabbar A, Almarshad F, Altherwy YN, Akram T. A review on automated plant disease detection: motivation, limitations, challenges, and recent advancements for future research. J King Saud Univ Comput Inf Sci. 2025;37(3):34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s44443-025-00040-3\u003c/span\u003e\u003cspan address=\"10.1007/s44443-025-00040-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeorge R, et al. Past, present and future of deep plant leaf disease recognition: A survey. Comput Electron Agric. 2025;234:110128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compag.2025.110128\u003c/span\u003e\u003cspan address=\"10.1016/j.compag.2025.110128\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMienye ID, Swart TG. A Comprehensive Review of Deep Learning: Architectures, Recent Advances, and Applications. Inf. 2024;15(12). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/info15120755\u003c/span\u003e\u003cspan address=\"10.3390/info15120755\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao X, Wang L, Zhang Y, Han X, Deveci M, Parmar M. A review of convolutional neural networks in computer vision. Artif Intell Rev. 2024;57(4):99. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10462-024-10721-6\u003c/span\u003e\u003cspan address=\"10.1007/s10462-024-10721-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRachmad A, Fuad M, Rochman EMS. Convolutional Neural Network-Based Classification Model of Corn Leaf Disease. Math Model Eng Probl. 2023;10(2):530\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.18280/mmep.100220\u003c/span\u003e\u003cspan address=\"10.18280/mmep.100220\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosadi MI, Lutfi M. Identifikasi Jenis Penyakit Daun Jagung Menggunakan Deep Learning Pre- Trained Model. J Explor IT. 2021;13(36):1\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePurnomo H, Mudjirahardjo P, Pramono SH. Implementation of CBIR Method for Identification of Corn Disease Using Extreme Learning Machine. Int J Comput Appl Technol Res. 2022;11(02):14\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7753/ijcatr1102.1002\u003c/span\u003e\u003cspan address=\"10.7753/ijcatr1102.1002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDame TA, Adera GB, Girmaw DW. Deep learning-based potato leaf disease classification and severity assessment. Discov Appl Sci. 2025;7(7):649. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s42452-025-07240-6\u003c/span\u003e\u003cspan address=\"10.1007/s42452-025-07240-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMyllyaho L, Raatikainen M, M\u0026auml;nnist\u0026ouml; T, Mikkonen T, Nurminen JK. Systematic literature review of validation methods for AI systems. J Syst Softw. 2021;181:111050. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jss.2021.111050\u003c/span\u003e\u003cspan address=\"10.1016/j.jss.2021.111050\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmin H, Darwish A, Hassanien AE, Soliman M. End-to-End Deep Learning Model for Corn Leaf Disease Classification. IEEE Access. 2022;10:31103\u0026ndash;15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/ACCESS.2022.3159678\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2022.3159678\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMehta AR, Kumar P, Prem G, Aggarwal S, Kumar R. AI-powered Innovations in Agriculture: A Systematic Review on Plant Disease Detection and Classification, \u003cem\u003eIndian J. Agric. Res.\u003c/em\u003e, no. Of, pp. 1\u0026ndash;11, 2025, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.18805/ijare.a-6371\u003c/span\u003e\u003cspan address=\"10.18805/ijare.a-6371\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalka TD, Hanafi MB, Rahman SMSAA, Zulperi DBM, Omar Z. Plant leaf disease detection and classification using convolution neural networks model: a review. Artif Intell Rev. 2025;58(10):322. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10462-025-11234-6\u003c/span\u003e\u003cspan address=\"10.1007/s10462-025-11234-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eG\u0026uuml;lmez B. A Comprehensive Review of Convolutional Neural Networks based Disease Detection Strategies in Potato Agriculture. Potato Res. 2025;68(2):1295\u0026ndash;329. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11540-024-09786-1\u003c/span\u003e\u003cspan address=\"10.1007/s11540-024-09786-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmad A, Gamal AE, Saraswat D. Toward Generalization of Deep Learning-Based Plant Disease Identification Under Controlled and Field Conditions. IEEE Access. 2023;11:9042\u0026ndash;57. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/ACCESS.2023.3240100\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2023.3240100\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarim MJ, Goni MOF, Nahiduzzaman M, Ahsan M, Haider J, Kowalski M. Enhancing agriculture through real-time grape leaf disease classification via an edge device with a lightweight CNN architecture and Grad-CAM. Sci Rep. 2024;14(1):16022. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-024-66989-9\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-66989-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKabel SAE-M, El-Shafai W, El-Samie FEA, Mohamed B. Enhanced plant identification and disease diagnosis through SqueezeNet and SVM for smart agriculture applications. Iran J Comput Sci. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s42044-025-00237-9\u003c/span\u003e\u003cspan address=\"10.1007/s42044-025-00237-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTunio MH, et al. Advancing plant disease classification: A robust and generalized approach with transformer-fused convolution and Wasserstein domain adaptation. Comput Electron Agric. 2024;227:109574. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compag.2024.109574\u003c/span\u003e\u003cspan address=\"10.1016/j.compag.2024.109574\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTimilsina S, Sharma S, Kondo S. Advancements in maize leaf disease detection, segmentation and classification: A review. Biosyst Eng. 2025;255:104162. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.biosystemseng.2025.104162\u003c/span\u003e\u003cspan address=\"10.1016/j.biosystemseng.2025.104162\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu W, Zhang A. Plant Disease Detection Algorithm Based on Efficient Swin Transformer. Comput Mater Contin. 2025;82(2):3045\u0026ndash;68. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.32604/cmc.2024.058640\u003c/span\u003e\u003cspan address=\"10.32604/cmc.2024.058640\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuhan S, Gulia P, Gill NS, Narwal E. RTR_Lite_MobileNetV2: A lightweight and efficient model for plant disease detection and classification. Curr Plant Biol. 2025;42:100459. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cpb.2025.100459\u003c/span\u003e\u003cspan address=\"10.1016/j.cpb.2025.100459\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhan AT, Jensen SM, Khan AR, Li S. Plant disease detection model for edge computing devices, \u003cem\u003eFront. Plant Sci.\u003c/em\u003e, vol. 14, no. December, pp. 1\u0026ndash;10, 2023, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fpls.2023.1308528\u003c/span\u003e\u003cspan address=\"10.3389/fpls.2023.1308528\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu X, Fan X, Luo P, Das Choudhury S, Tjahjadi T, Hu C. From Laboratory to Field: Unsupervised Domain Adaptation for Plant Disease Recognition in the Wild. Plant Phenomics. 2023;5. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.34133/plantphenomics.0038\u003c/span\u003e\u003cspan address=\"10.34133/plantphenomics.0038\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeng Y, Zhan J, Li K, Yan F, Zhang L. A rapid and precise algorithm for maize leaf disease detection based on YOLO MSM. Sci Rep. 2025;15(1):6016. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-025-88399-1\u003c/span\u003e\u003cspan address=\"10.1038/s41598-025-88399-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShafik W, Tufail A, Namoun A, De Silva LC, A. RA, Apong HM. A Systematic Literature Review on Plant Disease Detection: Motivations, Classification Techniques, Datasets, Challenges, and Future Trends, \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 11, no. May, pp. 59174\u0026ndash;59203, 2023, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/ACCESS.2023.3284760\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2023.3284760\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNgugi HN, Akinyelu AA, Ezugwu AE. Machine Learning and Deep Learning for Crop Disease Diagnosis, \u003cem\u003eAgronomy\u003c/em\u003e, vol. 14, no. 12, 2024, [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.mdpi.com/2073-4395/14/12/3001\u003c/span\u003e\u003cspan address=\"https://www.mdpi.com/2073-4395/14/12/3001\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlbert ETA, Bille NH, Leonard NME. Leveraging deep learning for plant disease identification: a bibliometric analysis in SCOPUS from 2018 to 2024. J Sci Agric. 2025;9:16\u0026ndash;39. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.25081/jsa.2025.v9.9412\u003c/span\u003e\u003cspan address=\"10.25081/jsa.2025.v9.9412\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHasan S, Jahan S, Islam MI. Disease detection of apple leaf with combination of color segmentation and modified DWT. J King Saud Univ - Comput Inf Sci. 2022;34(9):7212\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jksuci.2022.07.004\u003c/span\u003e\u003cspan address=\"10.1016/j.jksuci.2022.07.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbdou MA. Literature review: efficient deep neural networks techniques for medical image analysis, \u003cem\u003eNeural Comput. Appl.\u003c/em\u003e, vol. 34, no. 8, pp. 5791\u0026ndash;5812, 2022, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00521-022-06960-9\u003c/span\u003e\u003cspan address=\"10.1007/s00521-022-06960-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoihan A, Sunarya PA, Rafika AS. Pemanfaatan Machine Learning dalam Berbagai Bidang: Review paper. IJCIT (Indonesian J Comput Inf Technol. 2020;5(1):75\u0026ndash;82. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.31294/ijcit.v5i1.7951\u003c/span\u003e\u003cspan address=\"10.31294/ijcit.v5i1.7951\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQotrunnada FM, Utomo PH. Metode Convolutional Neural Network untuk Klasifikasi Wajah Bermasker. Prisma. 2022;5:799\u0026ndash;807.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArini A-, Wardhani LK, Octaviano D-. Perbandingan Seleksi Fitur Term Frequency \u0026amp; Tri-Gram Character Menggunakan Algoritma Na\u0026iuml;ve Bayes Classifier (Nbc) Pada Tweet Hashtag #2019gantipresiden. Kilat. 2020;9(1):103\u0026ndash;14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.33322/kilat.v9i1.878\u003c/span\u003e\u003cspan address=\"10.33322/kilat.v9i1.878\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAngga Irawan F, Sudarma M, Khrisne DC. Rancang Bangun Aplikasi Identifikasi Penyakit Tanaman Pepaya California Berbasis Android Menggunakan Metode CNN Model Arsitektur Squeezenet, \u003cem\u003eSpektrum\u003c/em\u003e, vol. 8, no. 2, pp. 18\u0026ndash;26, 2021, [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ojs.unud.ac.id/index.php/spektrum/article/view/75393\u003c/span\u003e\u003cspan address=\"https://ojs.unud.ac.id/index.php/spektrum/article/view/75393\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSetiawan W. Perbandingan Arsitektur Convolutional Neural Network Untuk Klasifikasi Fundus. J Simantec. 2020;7(2):48\u0026ndash;53. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.21107/simantec.v7i2.6551\u003c/span\u003e\u003cspan address=\"10.21107/simantec.v7i2.6551\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAngjaya HP, Gunadi K, Adipranata R. Pengenalan Penyakit pada Tanaman Pokok di Indonesia dengan Metode Convolutional Neural Network, \u003cem\u003eJ. Infra\u003c/em\u003e, 2021, [Online]. Available: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://publication.petra.ac.id/index.php/teknik-informatika/article/view/11426%0Ahttp://publication.petra.ac.id/index.php/teknik-informatika/article/viewFile/11426/10036\u003c/span\u003e\u003cspan address=\"http://publication.petra.ac.id/index.php/teknik-informatika/article/view/11426%0Ahttp://publication.petra.ac.id/index.php/teknik-informatika/article/viewFile/11426/10036\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVujović Ž. Classification Model Evaluation Metrics. Int J Adv Comput Sci Appl. 2021;12(6):599\u0026ndash;606. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.14569/IJACSA.2021.0120670\u003c/span\u003e\u003cspan address=\"10.14569/IJACSA.2021.0120670\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao J, Ding M, Sun Q, Dong J, Wang H, Ma Z. Classification of Southern Corn Rust Severity Based on Leaf-Level Hyperspectral Data Collected under Solar Illumination. Remote Sens. 2022;14(11):2551. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/rs14112551\u003c/span\u003e\u003cspan address=\"10.3390/rs14112551\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRozaqi AJ, Sunyoto A, Arief R. Deteksi Penyakit pada Daun Kentang Menggunakan Pengolahan Citra dengan Metode Convolutional Neural Network Detection of Potato Leaves Disease Using Image Processing with Convolutional Neural Network Methods. Creat Inf Technol J (CITEC JOURNAL). 2021;8(1):22\u0026ndash;31.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Convolutional Neural Network, Corn Leaf Disease Detection, Image Classification, Precision Agriculture, Deep Learning","lastPublishedDoi":"10.21203/rs.3.rs-7872919/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7872919/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The research develop an accurate and efficient method for detecting multiple corn leaf diseases to support sustainable agricultural practices in Soppeng Regency, Indonesia. The goal is to design a Convolutional Neural Network (CNN) model capable of classifying corn leaf diseases, including rust, blight, and gray leaf spot, using high-resolution image data. The research employed a balanced dataset sourced from open-access repositories, followed by preprocessing, data augmentation, and CNN model optimization. The model’s performance was evaluated using accuracy, precision, recall, and F1-score to ensure comprehensive assessment. Experimental results show that the proposed CNN achieved high accuracy across all disease classes, with strong per-class metrics, indicating robust performance in distinguishing visually similar symptoms. The classification results with the Convolutional Neural Network algorithm have 95% training data accuracy and 93% test data accuracy in detecting leaf diseases in corn plants. The findings contribute to agricultural technology by offering a scalable and field-deployable disease detection system that can be integrated into mobile or edge-based platforms. Limitations include reliance on publicly available datasets, which may not fully capture the variability of local field conditions. The research concludes that the proposed CNN model can significantly enhance early disease detection, reduce dependency on manual inspections, and support precision agriculture. Future research should focus on expanding the dataset with locally captured images, incorporating real-time image acquisition, and optimizing the model for deployment in low-resource environments to improve adaptability and reliability.","manuscriptTitle":"Deep Learning Based Multiclass Detection of Corn Leaf Diseases Using a Convolutional Neural Network","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-22 10:23:14","doi":"10.21203/rs.3.rs-7872919/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"92bf9a75-26ef-463f-9ce4-80daed4d15f8","owner":[],"postedDate":"December 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-18T09:11:17+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-22 10:23:14","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7872919","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7872919","identity":"rs-7872919","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}