Plant Leaf Disease Detection and Automatic Pesticide Recommendation Using Deep Learning

Abstract

Defoliating plant diseases are critical and still remains a significant threat on the world’s agricultural production,foodsecurityandeconomies.Theoldermodelsfor disease detection and diagnosis are inefficient,error-prone,and labor extensive, allowing for delays in action and mistimed pesticide application. Recent improvements in deep learning (DL)andimageprocessingtechnologiesappeartoallowtheuse ofautomated plant disease detectionsystems that canassist in recommending suitable and more effective pesticides. This work assesses various deep learning methods for their capabilities for plant leaf diseases detection, focusing on their functionality, accuracy and ease of use in real field setting. It also focuses on the methods for implementing the automatic recommendation systems interpolating between CNN and machine learning methodologies. It was shown that the most effective architectures are implemented on the basis of CNN, giving the best results in disease diagnosis precision. Also, the hybrid methods allowcombining the recommendation for the use of pesticides. A number of possible avenues for research aimed at enhancing efficiency of the models in real time field application are outlined.

Introduction

Agricultural productivity is crucial for global food security and economic stability, yet plant diseases cause significant crop losses (20-40% annually), impacting yield, quality, prices, and supply chains. Traditional disease detection relies on manual inspection of leaves, which is slow, error-prone, and impractical for large-scale farming. Incorrect pesticide use further harms crops and ecosystems.

Artificial Intelligence, especially deep learning models like Convolutional Neural Networks (CNNs), offers a promising solution by accurately identifying diseases from leaf images without manual feature extraction. CNNs have achieved over 90% accuracy in detecting diseases across crops such as corn, mango, and citrus.

Beyond detection, integrated systems that recommend appropriate pesticides based on the diagnosed disease are needed to minimize pesticide overuse and environmental damage. Hybrid models combining CNN-based detection with machine learning (ML) algorithms analyze disease severity and climate conditions to provide precise pesticide suggestions.

However, challenges remain in deploying these models in real-world farming due to computational demands, environmental variability (lighting, occlusions), and the diversity of diseases across crops and regions. Limited datasets hamper generalization across different conditions.

Research goals focus on evaluating CNN models for accuracy, efficiency, and scalability and integrating them with ML for automated pesticide recommendation. The literature review highlights:

• Early image processing methods, less accurate but computationally light.

• CNN-based architectures that improve detection accuracy.

• Lightweight CNN models (e.g., MobileNet) suitable for resource-limited environments, though sometimes less accurate.

• Hybrid ML-DL models that enhance pesticide recommendation specificity.

• Real-time, field-deployable systems that consider environmental factors and provide timely pesticide suggestions.

Comparative analyses show that deeper CNNs offer high accuracy but require significant computational power, while lightweight models trade some accuracy for efficiency. Robustness to field conditions and timely pesticide recommendation are key for practical agricultural application.

Conclusion

As a result of this research, existing frameworks have been evaluated and analyzed that focus on plant leaf microbialdisease detection and automatic pesticide recommendation through deep learning,theiressentialchanges,strengths,andweaknesses.Deep learning-based new generation systems, including convolutional neuralnetworks(CNN),havedrasticallyincreasedtheaccuracyof disease diagnosis, improving even the existing image processing techniques.EfficientandlightweightstructuresforMobileNetand EfficientNet have shown their potential for field applications whichrequirerealtimeoperationofresourceconstrainedsystems. ItcanbenotedthathybridmodelsofCNN-baseddiseasediagnosis andmachinelearningalgorithmsforrecommendationsystemsare moreversatileinnature,however,theyarestillaworkinprogress with regards to their applicability on new diseases and varied agriculturalpractices. In as much as the advances have been achieved, actual implementationinpracticalscenariosisstilladauntingtask.Many models exhibit reduced efficiency in field settings due to the presence of lighting differences, obstruction and the diversity of theleaves alone. Also, dataset diversity must be emphasized as a veryimportantaspectwhenseekingtodevelop intelligentmodels becauseasystemdesignedusingasmallsetofconditionswillnot be able to work over many crops, diseases and environments. Thesearethetwofactorswhichifincorporatedwillhelpcreatean umbrella model that will ultimately improve performance, reliability and increase usability in real world settings. Going forward there is a clear demand for systems that can improveon theexistinglimitations and that willmakeuseof low operational resources to detect diseases at high efficiency standards. The advancement of plant protection measures will be facilitated by extending the research on hybrid and multi-model strategies, developing extensive agricultural databases, and enhancing the abilitytocopewithchangesinrealtime.Suchadvancementsmay bebeneficialfordecreasingcroplosses,reducingthedependency on pesticides, and promoting agriculture9s sustainable development. The integration of these systems, therefore, can revolutionize agricultural disease management through deep learning for plant disease detection and pest control, ultimately transformingagriculturalproductivityandtheglobalenvironment for the better.

References

[1] \" Ahmad, M., Hussain, A., Basit, A., & Zafar, N. (2022). ACombined Architecture of Image Processing Techniques andDeep Neural Network for the Classification of Corn PlantDiseases. Journal of Plant Pathology, 125(4), 489-496. [2] Alaskar, H., & Al-Tuwaijri, S. (2022). A Mango DiseasePrediction for Smart Agriculture Using Machine LearningAlgorithms. IEEE Access, 10, 24502-24512. [3] Anand, S., & Mohan, K. (2021). A Review of Plant DiseaseDetection Methods Using Image Processing Approaches.InternationalJournalofComputerApplications,183(15),29-37. [4] Anitha, M., & Pushparaj, A. (2022). A Review on CoconutTree and Plant Disease Detection Using Various DeepLearningandConvolutionalNeuralNetworkModels.Journal of Agriculture andTechnology, 12(3), 123-135. [5] Ashok, B., & Kumar, C. (2021). A Review on Various PlantDisease Detection Using Image Processing. AppliedComputing and Informatics, 17(1), 121-134. [6] Baloch, A., & Khan, F. (2022). A Survey: Machine Learningand Deep Learning in Wheat Disease Detection andClassification.ComputersandElectronicsinAgriculture,189, 106358. [7] Bharathi, G., & Lakshmi, T. (2021). A Survey on AutomatedDisease Diagnosis and Classification of Herbal Plants UsingDigital Image Processing. Information Processing inAgriculture, 8(4), 593-602. [8] Chari, R., & Sharma, R. (2022). A Web-Based AgricultureRecommendation System Using Deep Learning for Crops,Fertilizers,andPesticides.SustainableComputing:Informatics and Systems, 36, 100697. [9] Das, P., & Mahato, S. (2022). Accurate Prediction of LeafDisease in Pepper Bell Using Deep Learning Algorithm.Environmental Technology & Innovation, 28, 102281. [10] Dutta,A.,&Pal,R.(2023).AdvancedCitrusDiseaseAnalysis Integrating ML and DL for Improved Identificationand Classification. Journal of Plant Pathology, 109(2), 303-312. [11] Eapen, T., & Prakash, A. (2021). An Automated and Fine-Tuned Image Detection and Classification System for PlantLeafDiseases.ComputersandElectronicsinAgriculture,185, 106084. [12] Faruqui, S., & Asif, M. (2022). Application of Deep CNN forImage-Based Identification and Classification of PlantDiseases. Artificial Intelligence in Agriculture, 7, 80-88. [13] Gautam,D.,&Singh,N.(2023).AutomaticDiagnosisofPlant Leaf Diseases Based on Deep Learning Approach.Environmental Science andPollutionResearch, 30, 4562-4571. [14] Halder,R., &Gupta,S. (2022). BananaPlantDiseaseDetection Using Image Processing. Journal of ImageProcessing and Pattern Recognition, 25(3), 300-308. [15] Hussain, Z., & Alam, M. (2021). Comparative Analysis ofSmart Pesticide Recommendation System Using ML & AI.Agricultural Informatics, 13(2), 154-165. [16] Islam,M.,&Ali,T.(2023).CottonLeafDiseaseClassification and Pesticide Recommendation. Computers inBiology and Medicine, 146, 105654. [17] James, S., & Michael, D. (2021). Detection of Defected MaizeLeaf Using Image Processing Techniques. Indian Journal ofAgricultural Sciences, 85(5), 643-647. [18] Kapoor, S., & Shinde, P. (2023). Detection of Diseases inCardamom Leaves by Using Digital Image Feature SelectionTechniques. Journal of Plant Protection Research, 63(1), 92-101. [19] Kaur,G.,& Gill,R.(2022). Developmentof Plant-LeafDisease Classification Model Using Convolutional NeuralNetwork. Computers and Electronics in Agriculture, 192,106612. [20] Kumar, V., & Raj, M. (2023). Disease Detection inPomegranate Plant Using Image Processing Techniques.Journal of Applied Horticulture, 25(1), 59-66. [21] Lee, J., & Park, S. (2022). Disease Detection of Citrus PlantsUsing Image Processing Techniques. Agricultural SciencesJournal, 18(4), 400-412. [22] Malik, F., & Sharif, S. (2021). Earlier Detection of PlantDisease and Recommending Pesticides Using ConvolutionalNeural Network. Sustainable Agriculture Reviews, 28, 181-195. [23] Meena, P., & Singh, D. (2022). e-Farmer: A Study of HowImageProcessingToolsMayBeUsedtoDetectPlantDisease. International Journal of Applied AgriculturalSciences, 9(1), 18-26. [24] Munshi, S., & Asghar, T. (2022). Effective Classification ofPlant Disease Using Image Processing and Machine Learning.Computers and Electronics in Agriculture, 188, 106377. [25] Patel, H., & Desai, J. (2023). FieldPlant: A Dataset of FieldPlant Images for Plant Disease Detection and ClassificationWithDeepLearning.AgriculturalInformatics,15(3),225-238. [26] Qureshi, U., & Ahmad, S. (2022). Grape Leaf DiseaseRecognition: A Deep Learning and Machine LearningTechniques Overview. Journal of Agricultural Informatics,12(1), 97-105. [27] Rajesh, B., & Thakur, M. (2021). Identification of VariousDiseases in Plant Leaves Using Image Processing and CNNApproach.AgriculturalInformaticsJournal,17(2),155- [28] [28]Rao, A., & Singh, T. (2023). Leaf Diseases Prediction, PesticidesRecommendationUsingDeepClassificationandPredictionUsingConvolutionalNeural Learning Techniques. Smart Agricultural Technology, 5,100243. [29] Sharma,A., &Gupta, R.(2022). Machine Learning-BasedPlant Disease Detection for Agricultural Applications: AReview.ComputersandElectronicsinAgriculture,189162. 106411. [30] Thakur,R.,&Verma,D.(2021).PaddyPlantDiseaseDetection,ClassificationandPredictionUsingConvolutionalNeuralNetwork.ComputersinBiologyandMedicine,129,104145.

Copyright

Copyright © 2025 Dr. Jayshree Pansare, Urja Awasarmol, Ayush Shastrakar, Sanket Niture. 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.