Plant Disease Detection Using InceptionV3

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 06 | Jun 2022 www.irjet.net p ISSN: 2395 0072

Plant Disease Detection Using InceptionV3

Mishalee Lambat1, Rushab Kothari2, Mitrup Kabi3, Tushar Mane4

1,2,3 Student, Dept. of Computer Science and Engineering, MIT ADT University, India 4Professor, Dept. Computer Science and Engineering, MIT ADT University, India ***

Abstract - Plants have evolved into a major source of energy and hence play a vital role in solving the global warming challenges. There are many types of diseases that are present in plants. To detect these diseases patterns are required to recognize. To achieve their goals, many of the methods that use this approach use digital image processing tools. There are many types of pattern recognition algorithms that give detection of disease with accuracy. In the existing work backpropagation and principal component analysis are used to detect plant diseases. These algorithms are learned from training supervision in neural networks. There is an issue with the accuracy of these algorithms. These algorithms are able to detect diseases in plants but not in an accurate way. So, to increase the accuracy of plant detection a new method will be proposed. This research work is a survey based on the data and information in these research papers mentioned below.

Key Words: Digital image processing, Neural Networks, Pattern Recognition algorithms, Inceptionv3, CNN, Disease Detection, Cotton Disease Detection.

1. INTRODUCTION

Plant diseases can badly affect a wide range of crops, posing a major threat to agricultural production. Manual disease identificationcanbetime consumingandlabor intensive.Oneofthemostsignificantfactorshinderingplantgrowthisdisease attack.Accordingtoadetailedagriculturalstudy,diseasescouldbeidentifiedmoreeasilyusingmachinelearningtechniques insteadofamanualmethod.Asaresult,picturesoftheaffectedleavescanbeidentifiedusingamachinelearningmethod.To processthepicturescapturedbyacamera,variousimageprocessingmethodswillbeused.Thesemethodswouldthenassistin plantdiseasedetection,resultinginincreasedplantyield.Thisresearchpaperexaminesvariousdiseaseidentificationand classificationtechniquesaswellasplantdiseaserecognitionusingaMachineLearningapproach.

AgricultureisvitaltotheeconomiesofdevelopingcountriessuchasIndia.Plantdiseasehasconcentratedthequantityandthe quality of agricultural merchandise. Cotton is a major source of revenue in India. Cotton crops suffer once leaves fall off prematurelyor becomeinfected withdiseases.Formillennia,farmersand planting expertshavehad manyconcernsand ongoing agricultural obstacles, including cotton disease. Even though thesevere cotton disease can outcome in no grain harvest,aquick,efficient,low cost,anddependablemethodforidentifyingcottondiseasesishighlydesiredintheagriculture field.

Microorganisms'lifecyclescannotbepredicted.Somediseasesarenotvisibleintheearlystagesandonlyappearattheend. Plantdiseasepredictionbythenakedeyeisusedinpractice,buttheresultsaresubjecttointerpretation,anddiseaseextent isn'treallystrategicallyplaced.Nowadays,automaticidentificationofplantdiseaseisahotresearchtopic,andthusdetects diseasesbasedonsymptomsthatappearonplantleaves.Manyimageprocessingmethodshavebeenintroducedtoresolve issuesbypatternrecognitionandalsosomeautomaticsegmentationtools,dependingontheapplications.Inthenextsection, thesepaperspresentasurveyofthoseproposedsystemsinameaningfulway.Acomparisontableofvariousalgorithms, methodology,andtheiraccuracyhasbeenmadeasfollows:

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 06 | Jun 2022 www.irjet.net p ISSN: 2395 0072

2. LITERATURE SURVEY

Title Author Method Used Result Journal / Confere nce

Year of Publica tion

SCI/Scopus/IE EE

Deeplearning Konstantinos CNN Thefinalmodelachieved Journal 2018 IEEE modelsforplant Ferentinos 99.53%accuracyon17,548 diseasedetection previouslyunseenimages. anddiagnosis

PlantDisease EmmaHarte CNN, Theproposedmodelcan Journal 2020 IEEE Detectionusing RestNet34 obtainanaccuracyof97.2% CNN andanF1scoreofmorethan 96.5%,accordingto validationdata.

Riceplant VimalK. Support Usingasupportvector Journal 2020 SpringerLink disease Shrivastava & Vector machine(SVM)classifier,a classification MonojK. Machine performanceof94.65%was usingcolor Pradhan obtained. features:a machinelearning paradigm

AnOverviewof Ms.KiranR. CNN Themodel3D CNNhadan Journal 2014 IOSRJournal of theResearchon Gavhale, Prof. accuracyrateof95.73%. Computer

PlantLeaves Ujwalla Engineering Diseasedetection Gawande (IOSR JCE) usingImage Processing Techniques.

PlantDisease LiliLi, CNN, Theoveralldiseaseaccuracy ratewas98%attheleaflevel and94percentatthepixel level,accordingtothestudy.

Journal 2021 IEEE Detectionand Shujuan ResNet, Classificationby Zhang,Bin VGG16 DeepLearning Wang

PlantDisease Detectionand Classification usingCNN

RinuR, ManjulaS H

The technologie susedare deep

Theaccuracyachievedis around94.8%. Journal 2021 IRJET learning, CNN and VGG 16 Model.

ASurveyon SnehaPatel CNN Anaccuracyof93.82%was Journal 2020 IJMTST PlantLeaf ,Dr.U.K. achievedonPlantVillage Disease Jaliya, Detection Pranay Patel

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal

International Research Journal of Engineering and Technology (IRJET)

e ISSN: 2395 0056

Volume: 09 Issue: 06 | Jun 2022 www.irjet.net p ISSN: 2395 0072

Detectionof Diseaseson CottonLeaves UsingK Mean Clustering Method

Plantdisease detectionusing CNN

PlantDisease Detectionand ClassificationUsing DeepNeural Networks

PawanP. Warne,Dr.S. R.Ganorkar

K mean clustering algorithm

K MeanClustering algorithmhashighest accuracyof80.56%

SumitKumar, Veerendra Chaudhary,Ms .Supriya Khaitan Chandra

Aravindhan Venkatarama nan,Deepak KumarP Honakeri, Pooja Agarwal

CNN K MeanClustering algorithmhashighest accuracyof80.56%

CNN, Kerasdeep learning framework, ResNet

Aplantdisease identificationsystem withanaccuracyof96% wasdeveloped.

Real Time DetectionofApple LeafDiseases UsingDeep LearningApproach BasedonImproved Convolutional NeuralNetworks

ADeepLearning basedApproach forBananaLeaf Diseases Classification

UsingDeep Learningfor Image Based PlantDisease Detection

PengJiang, YuehanChen, BinLiu DongjianHe And Chunquan Liang

JihenAmara, Bassem Bouaziz,and Alsayed Algergawy

Sharada Prasanna Mohanty, DavidHughes andMarcel Salathé

Deep CNNs

Theexperimentalresults showthattheINAR SSD modelrealizesadetection performanceof78.80%

Journal 2015 IRJET

Detectionand classificationof riceplant diseases

Harshad kumarB. Prajapati,Jite shP. Shah,VipulK. Dabhi

Deep learning

Deep learning

Thetrainedmodelachieves anaccuracyofalmost98%

Journal 2021 Turkish Journalof Computer and Mathematic sEducation

Journal 2019 ISSN

K means clustering

Thetrainedmodel achievesanaccuracyof 99.35%onaheld out testset,demonstrating thefeasibilityofthis approach

Onthetrainingdataset, theywere93.33percent accurate,andonthetest dataset,theywere73.33 percentaccurate They alsoconducted5 and 10 foldcross validations,achievingan accuracyof83.80%and 88.57%,respectively.

Journal 2019 IEEE

Confer ence Paper

2017 IEEE

Journal 2016 Frontiersin PlantScience

Journal 2017 Intelligent Decision Technologies ,IOSPress

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 06 | Jun 2022 www.irjet.net p ISSN: 2395 0072

3. PROPOSED METHOD

TheproposedmethodusestheCNNimageclassificationtechniquetoclassifytheleafpicture.Basedonretrievedinformationat eachconvolutionlayer,itcandetectandrecognizediseasesautomatically.Thearchitecturalsystemoftheproposedsystemis depictedinthediagrambelow.Fordiseaseidentification,thesystemusedimageprocessingtechniques.Theusershouldfirst upload an image of a cotton plant leaf. The algorithm can preprocess the image before applying the CNN algorithm. The network was built using a combination of the benefit of pre trained on ImageNet and the Inception component, and this technique outperforms other state of the art techniques. Every convolution layer within dense block is tiny, so each convolutionkernelisstillinchargeoflearningthetiniestdetails.Thebelowimagesshowtheblockdiagramandflowchartof ourproject:

Fig -2:Flowchart

4. IMPLEMENTATION

Fig -3:BlockDiagram

I.DatasetCollection:TheDatasetwastakenfromtheKaggleofCottonDiseaseDatasetexistingonlineassuchthecodewasalso inscribedontheonlinekernelofKaggleforwellcomputationandstudyoftraininglossandvalidation.

II.DataPre Processing:Pre processingoftheinputimagetoimprovethequalityoftheimageandtoremovetheundesired distortionfromtheimage.Clippingoftheleafimageisperformedtogettheinterestingimageregionandthenimagesmoothing isdoneusingthesmoothingfilter.Toincreasethecontrast,Imageenhancementisalsodone.

II.Transferlearningisamethodinwhichinformationobtainedwhileattemptingtosolveoneproblemissavedandappliedto anotherrelatedproblem.ThemostcomplexpartofaCNNistypicallythefeatureextractionpart,whichisaveryissuethat emergeswithlimitedcomputationalresources.Byreusingthisextractionoffeaturessection,transferlearningmakesitsimple. Usersconsiderremovingthefinallayersofthepre trainedmodelandreplacingthemwithnewlayerstocreatethenewCNN. Thetrainingisonlydoneonthenewlyaddedlayers.

III.RecognitionBasedonInceptionV3NetworkTransferLearning:TheInceptionseriesofconvolutionalneuralnetworksisa seriesofneuralnetworksthatcannotbeignoredinthehistoryofconvolutionalneuralnetworks.Mostneuralnetworksonly deepenthedepthofthenetworkbyincreasingtheconvolutionallayertogetbetterperformancebeforetheemergenceofthe Inceptionneuralnetwork.Inceptionneuralnetworkshavechangedthisstrategy.TheInceptionmoduleproposedbyInception NeuralNetworkusesdifferentsizesoffiltersandmaximumpoolingtoreducethedimensionofthedata.Thishastheadvantage ofobtainingricherfeatureswithsignificantlyreducedcomputationandfewerparameters.

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal |

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

5. RESULT AND ANALYSIS

Fig -3:InceptionModel

TheabovefigureshowsTrain ValAccuracy.

6. CONCLUSIONS

TheabovefigureshowsTrain-ValLoss.

Thisresearchpaperdescribesastudyonvariousmethodsfordetectingplantleafdiseaseusingimageprocessingtechniques.A singlemethodcannotbeusedtoidentifyalldiseases.Manymethodshavebeendevelopedforidentifyingandclassifyingplant diseases using thediseased plant leaves dataset. However, there is currently no reliable and cost effective commercial technologyfordiseasedetection.Inourresearch,weusedCNNandInceptionModeltodetectonlycottondisease.Wehave usedtheKaggledatasetwithnearly500photostakeninlaboratoryconditionstotrainandtestthemodel.Furtherinthefuture, weintendtoexpandthisprojectbycreatingadatabaseofdifferentplantsandpredictingtheirdiseases.

ACKNOWLEDGEMENT

WewouldliketoexpressourspecialthanksandgratitudetoourguideProf.TusharMane,aswellasourPrincipalMr.Ravande whogaveusthegoldenopportunitytodothiswonderfulProjectontheTopicofPlantDiseaseDetection,whichhashelpedus in doing a lot of research and we came to know about so many new things about the Plant diseases, different types of algorithms,theirperformanceandaccuraciesthatareusedindetectingdiseases.

REFERENCES

[1]Shrivastava,VimalK.,andMonojKPradhan.n.d.“Riceplantdiseaseclassificationusingcolorfeatures:amachine learningparadigm.”JournalofPlantPathology103.1(2021):17 26.

[2]EmmaHarte.n.d.“PlantDiseaseDetectionusingCNN.”Diss.Ph.D.thesis,2020.

[3]Ferentinos,KonstantinosP.n.d.“Deeplearningmodelsforplantdiseasedetectionanddiagnosis.”Computersand ElectronicsinAgriculture145(2018):311 318.

[4]Gavhale,MsGawande,,andUjjwala.2014.“AnOverviewoftheResearchonPlantLeavesDiseaseDetectionusing ImageProcessingTechniques.”IOSRJournalofComputerEngineering16.10 16.10.9790/0661 16151016.

Volume: 09 Issue: 06 | Jun 2022 www.irjet.net p ISSN: 2395 0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal |

International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056

Volume: 09 Issue: 06 | Jun 2022 www.irjet.net p ISSN: 2395 0072

[5]L.Li,S.Zhang,andB.Wang.n.d.“PlantDiseaseDetectionandClassificationbyDeepLearning AReview.”IEEEvol.9, pp.56683 56698,2021(doi:10.1109/ACCESS.2021.3069646.).

[6]R.R,andM.SH.n.d.“PlantDiseaseDetectionandClassificationusingCNN.”InternationalJournalofRecentTechnology andEngineering(IJRTE)vol.10,no.3,pp.152156,Sep.2021.

[7]Patel,Sneha,Jaliya,andDrPatelPranay.n.d.“ASurveyonPlantLeafDiseaseDetection.”InternationalJournalfor ModernTrendsinScienceandTechnology6.129 134.10.46501/IJMTST

[8]Warne,PawanP,andSRGanokar.2015.“DetectionofDiseasesonCottonLeavesUsingK MeanClusteringMethod.”

[9]SumitKumar.n.d.“PlantDiseaseDetectionUsingCNN.”TurkishJournalofComputerandMathematicsEducation (TURNCOAT)12.12(2021)(2106 2112).

[10]Venkataramanan, Aravindhan, and Pooja. 2019. “Plant Disease Detection and Classification Using Deep Neural Networks.”

[11]Jiang,Peng,YuehanChen,BinLiu,DongjianHeandChunquanLiang.n.d.“Real TimeDetectionofAppleLeafDiseases Using Deep Learning Approach Based on Improved Convolutional Neural Networks.” IEEE Access 7 (2019) (59069 59080).

[12]Amara,Jihen,Bouaziz,Bassem,andAlgergawy,Alsayed.2017.ADeepLearning basedApproachforBananaLeaf DiseasesClassification79 88.

[13]Mohanty,Sharada,Hughes,David,andSalathe,Marcel.2016.“UsingDeepLearningforImage BasedPlantDisease Detection.”FrontiersinPlantScience7.10.3389/fpls.2016.01419.

[14]Prajapati,Harshadkumar,Shah,Jitesh,andDabhi,Vipul.n.d.“Detectionandclassificationofriceplantdiseases.” IntelligentDecisionTechnologies11.357 373.10.3233/IDT 170301.

[15]GaganpreetKaur,Mrs.,Kaur,Sarvjeet,andKaur,Amandeep.1September2018.“PlantDiseaseDetection:aReviewof CurrentTrends.”InternationalJournalofEngineering&TechnologyVolume7Number3.34.

2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal