Soybean Leaf Disease Detection Using Convolutional Neural Network

Abstract

Soybean is one of the most important agricultural cropsworldwideandservesasamajorsourceoffood,edible oil, and animal feed. The productivity and quality of soybean cropsaresignificantlyaffectedbyvariousleafdiseasessuch as Caterpillar damage, Rust, Bacterial Blight, and Diabrotica Speciosa. These diseases reduce crop yield and negatively impact agricultural production and the economic condition of farmers. Therefore, early and accurate disease identification is essential foreffectivecropmanagementandimprovedproductivity.Tradi-tionaldiseasedetectionmethodsmainlyrelyonmanualinspection by agricultural experts, which is time-consuming, costly, less efficient for large-scale farming, and prone to human error. This paper presents an automated soybean leaf disease detec-tionsystemusingimageprocessinganddeeplearningtechniques. The proposed system utilizes Convolutional Neural Networks (CNN) for automatic feature extraction and classification of healthyanddiseasedsoybeanleaves.ARandomForestclassi-fierisalsoimplementedforcomparativeanalysistoevaluate the performance of traditional machine learning approaches.The collected dataset undergoes preprocessing steps such as image resizing, normalization, filtering, and data augmentation techniques including rotation, zooming, and flipping to improve model performance and reduce overfitting. Experimental results demonstrate that the CNN-based model achieves higher accuracy, precision, recall, and F1-score com-paredtoRandomForestmethods.Theproposedsystemprovides an efficient, reliable, and cost-effective solution for real-time soybean leaf disease detection. It can support smart agriculture systems and assist farmers in taking timely preventive measures to minimize crop damage and improve overall agricultural productivity.

Introduction

Soybean is an economically important oilseed crop widely cultivated in India due to its high protein content and industrial value. However, soybean production is significantly affected by leaf diseases such as Rust, Bacterial Blight, Caterpillar Damage, and Diabrotica Speciosa, which reduce crop quality and yield. Early disease detection is essential because delayed diagnosis can lead to severe economic losses for farmers.

Traditional disease detection methods rely on manual visual inspection by farmers or agricultural experts. These methods are time-consuming, labor-intensive, subjective, and often inaccurate, especially in large farming areas. Limited access to agricultural specialists in rural regions further complicates timely disease identification.

To overcome these limitations, this study proposes an automated soybean leaf disease detection system using image processing and Convolutional Neural Networks (CNNs). The objective is to accurately classify healthy and diseased soybean leaves while comparing CNN performance with the Random Forest algorithm for real-time agricultural applications.

The dataset consists of soybean leaf images belonging to five classes: Healthy, Rust, Bacterial Blight, Caterpillar Damage, and Diabrotica Speciosa. Images were collected under varying lighting conditions, backgrounds, and leaf orientations to improve model robustness. Preprocessing steps included image resizing, normalization, noise reduction, and data augmentation techniques such as rotation, flipping, and zooming to enhance dataset diversity and prevent overfitting.

The CNN model automatically extracts important features such as color variations, texture patterns, edges, spots, and disease-specific characteristics without requiring manual feature engineering. The architecture includes an input layer, convolutional layers, ReLU activation functions, max-pooling layers, fully connected layers, and a Softmax output layer for multi-class classification. Training was performed using labeled data with cross-entropy loss and the Adam optimizer.

The system workflow involves image acquisition, preprocessing, feature extraction, CNN-based classification, and result prediction through a web-based interface. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics.

Experimental results demonstrated that the proposed CNN model achieved 97–99% classification accuracy, significantly outperforming the Random Forest model, which achieved 85–88% accuracy. CNN showed superior precision, recall, and F1-scores due to its ability to automatically learn complex image features and effectively handle variations in lighting and background conditions. In contrast, Random Forest required manual feature extraction and struggled with complex disease patterns.

The proposed system offers several advantages, including automated disease detection, high accuracy, real-time prediction capability, reduced manual effort, and practical usefulness for farmers and agricultural experts. Overall, the study concludes that CNN-based deep learning provides a reliable and efficient solution for soybean leaf disease detection, helping improve crop management, reduce losses, and support precision agriculture.

Conclusion

This paper presents an automated soybean leaf disease detectionsystemusingConvolutionalNeuralNetworks(CNN) and image processing techniques. The proposed model effec-tivelyidentifieshealthyanddiseasedsoybeanleaveswithhigh accuracy,precision, andreliability.Experimental resultsshow that the CNN-based approach performs better than traditional machine learning methods such as Random Forest due to its automatic feature extraction capability. Thesystemreducesmanualeffortandsupportsearlydisease detection,whichcanhelpfarmerstaketimelypreventivemea-sures and reduce crop loss. The proposed method is efficient, cost-effective, and suitable for real-time smart agriculture applications. In the future, the system can be integrated with mobile and IoT-based platforms for practical field usage.

References

[1] M.A.Jahin,S.Shahriar,M.F.Mridha,J.Hossen,andN.Dey,“SoybeanDisease Detection via Interpretable Hybrid CNN-GNN: IntegratingMobileNetV2 and GraphSAGE with Cross-Modal Attention,” arXivpreprint, Mar. 2025. [2] S.Kaur,A.Pandey,andS.Goel,“Semi-automaticleafdiseasedetectionand classification system for soybean,” IET Image Processing, vol. 12,no. 5, pp. 1038–1048, 2021. [3] R.Bhavani,“DetectionofLeafDiseasesinSoybeanPlantusingAutoen-coderandMultinomialLogisticRegression,”LegumeResearch,vol.48,no. 5, pp. 876–884, 2025. [4] X.Peng,L.Wang,andY.Zhang,“EnhancedDetectionofSoybeanLeafDiseases Using an Improved YOLOv5 Model,” International Journal ofKnowledge and Innovation Studies, vol. 2, no. 1, pp. 45–56, 2024. [5] “Deep Learning Model for Classifying and Evaluating Soybean LeafDisease Damage,” International Journal of Molecular Sciences, vol. 25,no. 1, pp. 106–118, 2024. [6] A. Mohanty, D. Hughes, and M. Salathe´, “Using Deep Learning forImage-Based Plant Disease Detection,” Frontiers in Plant Science, vol.7, pp. 1419–1425, 2016. [7] S.Sladojevic,M.Arsenovic,A.Anderla,D.Culibrk,andD.Stefanovic,“Deep Neural Networks Based Recognition of Plant Diseases by LeafImage Classification,” Computational Intelligence and Neuroscience,vol. 2016, pp. 1–11, 2016. [8] K. P. Ferentinos, “Deep Learning Models for Plant Disease DetectionandDiagnosis,”ComputersandElectronicsinAgriculture,vol.145,pp.311–318, 2018. [9] J.G.A.Barbedo,“FactorsInfluencingtheUseofDeepLearningfor Plant Disease Recognition,” Biosystems Engineering, vol. 172, pp.84–91, 2018. [10] P.Sharma,Y.Hans,andS.Gupta,“ClassificationofPlantLeafDiseasesUsing Machine Learning and Image Processing,” International Journalof Engineering Research and Technology, vol. 9, no. 6, pp. 234–240,2020. [11] R. Brahimi, K. Boukhalfa, and A. Moussaoui, “Deep Learning forTomato Diseases: Classification and Symptoms Visualization,” AppliedArtificial Intelligence, vol. 31, no. 4, pp. 299–315, 2017. [12] S. Too, L. Yujian, S. Njuki, and Y. Yingchun, “A Comparative StudyofFine-TuningDeepLearningModelsforPlantDiseaseIdentification,”Computers and Electronics in Agriculture, vol. 161, pp. 272–279, 2019. [13] M. Picon, A. Alvarez-Gila, A. Seitz, et al., “Deep Convolutional NeuralNetworks for Mobile Capture Device-Based Crop Disease Classifica-tion,” Agronomy, vol. 9, no. 1, pp. 3–15, 2019. [14] H. Durmus, E. Gunes, and M. Kirci, “Disease Detection on the Leavesof the Tomato Plants by Using Deep Learning,” IEEE InternationalConference on Agro-Geoinformatics, pp. 1–5, 2017. [15] S. Ramesh, D. Vydeki, and K. Venkataraman, “Plant Disease DetectionUsing Machine Learning,” International Conference on Design Innova-tions for 3Cs Compute Communicate Control, pp. 41–45, 2018. [16] A. Ramcharan, K. Baranowski, P. McCloskey, et al., “Deep LearningforImage-BasedCassavaDiseaseDetection,”FrontiersinPlantScience,vol. 8, pp. 1852–1860, 2017. [17] T. Rahman, M. K. Hasan, and M. S. Islam, “CNN-Based Plant LeafDisease Detection Using Image Processing,” Journal of Artificial Intel-ligence and Capsule Networks, vol. 2, no. 1, pp. 5–14, 2020. [18] N. P. Patel and D. Shah, “Soybean Leaf Disease Detection UsingConvolutional Neural Networks,” International Journal of ComputerApplications, vol. 183, no. 12, pp. 20–25, 2021. [19] V. Singh and A. Misra, “Detection of Plant Leaf Diseases Using ImageSegmentationandSoftComputingTechniques,”InformationProcessingin Agriculture, vol. 4, no. 1, pp. 41–49, 2017. [20] S. Pawar and M. Kokare, “Leaf Disease Detection Using FeatureExtraction and Neural Networks,” International Journal of AdvancedComputer Research, vol. 5, no. 2, pp. 115–120, 2019. [21] R. Sujatha and Y. Chatterjee, “A Deep Learning-Based Approach forPlant Disease Detection,” Journal of Physics: Conference Series, vol.1717, no. 1, pp. 1–8, 2021. [22] P. Revathi and M. Hemalatha, “Classification of Cotton Leaf SpotDiseases Using Image Processing Edge Detection Techniques,” IEEEInternational Conference on Emerging Trends in Science, Engineeringand Technology, pp. 169–173, 2012.

Copyright

Copyright © 2026 Aarti Pandurang Chavare, Ajit B Patil, Tanvi Gopinath Patil, Dipti Prakash Swami, Manasi Dipak Shirale. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.