Plant Disease Detection Using CNN Model And Image Processing - IJERT

1y ago
641.68 KB
7 Pages
Last View : 15d ago
Last Download : 4m ago
Upload by : Amalia Wilborn

Published by :http://www.ijert.orgInternational Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020Plant Disease Detection using CNN Model andImage ProcessingMd. Tariqul IslamDepartment of Electronics and Communication Engineering (ECE)Khulna University of Engineering and Technology (KUET)Abstract— The rate of plants and crops cultivation ratesgrowing rapidly with the increment of human and animaldemands all over the world. The agricultural science inventedlots of authentic techniques to use in cultivation sector forimprove the production rate. During cultivation, farmers facedlots of challenges to protect their plant form different diseasesand insects which make their production lower and they facedmuch financial losses. Form decades, agricultural scientists triedhardly to invent a quick medication system for detect plantdisease quickly and gives treatment immediately. But theidentification techniques of fetal abnormalities of plant aremanual and it takes huge time. For reducing detection time andincreasing efficiency of plant disease detection lots of newtechnologies introduced with cultivation system. In this work weintroduced a model with the help of computer science andengineering using machine learning specially deep learning fordetecting the leaf disease by the image of “Corn”, “Peach”,“Grape”, “Potato” and “Strawberry”. In Bangladesh, Mize andPotato is very popular food item and Strawberry is also veryappealing for all aged people. Peach and Grape is not so popularcultivation in our country but their attraction in fruit marketskyrocketed. The farmer who is interested and involved tocultivate corn, peach, grape, potato and strawberry faces lots ofdisorder and attract of insects and huge loss faced by thefarmer. For prevent these losses and provide immediate cure, inthis paper uses image processing, CNN (Convolutional NeuralNetwork) model for training the dataset. Our system providesaccuracy rate 94.29% successfully. The universal cultivator canget assistance from our research work for growing theproduction rate of crops and fruits and alleviate the plantdisease and insect attract.maize act as a good source of minerals, dietary fiber andvitamins. There exist different types of corn for instance popcorn, dent corn, flour corn, sweat corn and flint corn. Springseason consider as the best time period for maize plantationand the corn is unable to tolerate coolness. The plant growsrapidly with the moisture soil. The cultivator faced differenttypes of critical challenges for protecting maize from differenttypes of disease which reduces the production rate. There arelots of maize diseases for example Anthracnose, Black bundleand Late wilt, Charcoal-Rot, Common Rust, Downy mildews,Cercospora leaf spot, Common sumt (Boil sumt, Blister sumt),Giberrella stalk and ear rot, Northern leaf blight , Southerncorn leaf blight, Bacterial leaf blight, Bacterial leaf streak,Bacterial stalk rot, Goss’s bacterial blight, Holcus spot,Stewart’s walt, Maize dwarf mosaic, pythium root rot, slugs,Aphids, Corn earworm, Cutworms, Fall armyworm, Fleabeetles, Thrips, Root knot nematode, Spider mites and so on.During Black bundle and Late wilt disease, the infected plantshows symptom after reaching tassel state. When this diseasestarted its attack the top leaves whose color is dull green andlosing its color gradually and finally dry. This disease can becontrolled by altering the crops, treatment the seed andreducing water stress. The top and bottom part of leavesbecome golden brown powdery color in common rust diseasesand this disease can be controlled by hybrid the plant. At thetime of growing, the leaves shows yellow rings brown spot incercospora leaf spot and the disease can be reduced by rotateand hybrids the plant.Keywords— Machine learning, CNN, Image Processing,Computer Vision.The peach is very popular fruits and very testy for eatingcontaining vitamin A. The peach grows in warmertemperature in the hemispheres of Northern and Southernregion. Peach is first invented in China after that it spreadsAsia, Europe, Spain, Mexico and United States. Peach (Prunuspersica) fruit tree size is very short. The peach’s cultivatorpaced different types of disease which reduced the peachproduction rate and causes huge losses. There are numerousdiseases such as Bacterial canker (Pseudomonas syringae),Bacterial spot (Xanthomonas campestris), Crown gall(Agrobacteium spp), Scab (Cladosporium carpophlum),Brown rot (Monilinia fructicola), Rust (Tranzschelia discolor),Short hole disease (Wilsonomyces carpophilus), Silver leafdisease (Chondrosterum pupureum), Leaf curl (Taphrinadeformans), Plum pox virus (PPV), Fruittree leafroller(Archips argyrospila), Oriental fruit moth (Grapholithamolesta) and so on. During Bacterial spot disease leaves underside become purple color with shot hole at the center and dropout the leaf from the tree. When peach faced Scab disease thefruits surface introduce small circular spot with green colorand gradually its size rising become dark produced yellowI.INTRODUCTIONBangladesh is an agricultural country. Around 80 percentpeople directly or indirectly related with the agronomicalservices. Bangladeshi economy extremely depends onagricultural department and large portion of economy inBangladesh comes from this sector. Our financial status ismoved forward by cultivating crops and fruits yearlong. InBangladesh numerous crops are cultivated over the countryand among all the crops and fruits rice, wheat and potatotouched the height position of their popularity. The plantationof Maize, Peach, Grape and Strawberry shows increasingpattern and the invitation of this cultivation enhance rapidly.So the interest of farmer for cultivating these sorts of cropsand fruits shows higher than any preceding decay.Zea mays is known as the scientific name of Maize and it’sanother name is corn or mielie. Around the world the corn isthe most cultivated crops and for numerous uses corncultivated commercially almost every country. People ofdifferent culture and nature are enjoyed corn as food and work is licensed under a Creative Commons Attribution 4.0 International License.)291

Published by :http://www.ijert.orgInternational Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020halo. In Brown rot disease the skin and tissue of fruits lossesits color. The peach upper and lower side shows angular shapeyellow green spots. The young leaf turn into red color fromyellow color and the leaf raised irregularly is known as leafcurl disease.Grapes is very testy fruits with green, red, purple color,seedless grapes, jelly grapes, jam grapes, grapes juice and soon. Grapes holds lots of vitamin A and B. Grapes can be eatenas cure of different jeopardy diseases such as Diabetes, eyeproblems, cardiovascular disease, cancer, heart disease, highblood pressure and so on. During cultivation grapes lots ofdisease are faced by the plant and fruits such as Powderymildew (Uncinula necator), Downy mildew (Plasmoparaviticola), Anthracnose (Elsinoe ampelina), Black rot(Guignardia bidwellii), Bacterial canker ( Xanthomonascampestris pv. Viticola), Brown leaf spot, Rust,Coniothyrium blight, Alternaria blight, Drechslera leaf spotand so on. Powdery mildew is very dangerous disease. Thisdisease attacks the vines and aerial part of the plant and firstlycluster and berry infection arrives. The leaf experiencedcircular spot which is brown color known as Black rot diseaseand the disease can be controlled by separating attacked fruits.The gray mold or bunch rot disease attacked the flowers andripen fruits. Potato is very popular food item around the worldand it act as vegetable. It has large leaf of small plant and thepotato produces under the soil. Potato can be grown in highand cool region. Potato is helpful for different disease likeblood pressure, brain functioning and nervous system,immunity, inflammation, digestion, heart health, skin care,cancer risk and so on. Potato production are reduced bydifferent types of diseases such as Common scab, Powderyscab, Rhizoctonia, Silver scurf, Bacterial spot rot, Blackleg,Early blight, Freezing and chilling injury, Fusarium dry rot,Late blight, leak, Mechanical injury and cracking, Pink rot,Ring rot, Root knot nematode, Blackheart, Black spot,Fusarium wilt, Net necrosis, Verticillium wilt and so on.Strawberry is very popular and testy fruit all over theworld and its family is Rosaceae, genus Fragaria. Strawberryis cultivated all over the world and mostly grew in theNorthern Hemisphere temperate regions. Strawberry is firstinvented in Europe then it is spreads all over the world and it’scommercial production have two format for consumptionimmediate and processing. Strawberry is available duringsummer and it is popular nutritious fruits with antioxidantcontent. Strawberry can prevent lots of disease such as Heartdisease, stroke, cancer, Blood pressure, Constipation,Diabetes. Strawberry is full of vitamin C, potassium, fiber andfoli acid. Strawberry plant and fruits are attacked by numerousdiseases for instance Angular leaf spot (Xanthomonasfragariae), Leaf scorch (Diplocarpon earlianum), Anthracnose(Colletotrichum fragariae), Gray mold (Botrytis cinerea), Leafspot (Mycosphaerella fragariae, Phomopsis leaf blight(Phomopsis obscurans), Powdery mildew (Spaerothecamacularis), Red stele or Red core (Phytophthora fragariae),Slugs, Aphids, Armyworm, Japanese beetle, Loopers and soon. During Angular leaf spot disease the lower side of leavesshow little water soaked which become large gradually andturn onto dark green color or angular spots. For protecting thisIJERTV9IS100123disease need to crop rotation, avoiding over irrigation,controlling over chemical use. The upper leaf side introduceddark or brown spot due to leaf scorch disease. This diseasescan be controlled by renewing regular plants, well aircirculating area choose for plantation, well drainage systemect.In Bangladesh major portion of cultivator are illiterate andthey are unable to detect disease using recent technologies. Inour country, the new invented algorithm CNN (ConvolutionalNeural Network) is not used by our farmer. Our cultivator isused hand-made, non-scientific techniques for harvestingcrops and detection of diseases. For protecting these diseasesthey preferred to use pesticides without appropriate scalewhich damages natural ecosystem. Most of them are largelydepend on eye view or blind guesses for disease detectionwhereas USA, China developed country used various moderntechnologies like CNN, AI and mostly image processingtechniques to detect or harvest their crops. In our researchwork we developed a model using machine learning so thatour farmer can easily use this technique to identify and giveappropriate cure. For training and testing our dataset throughdeveloping model here used CNN algorithm.II. LITERATURE REVIEWBefore doing this work, we read and try to understandsome source paper work so that we can do our workaccurately. The paper which we read before starting this workis introduced here as literature review. During reviewing thesepapers it is clear that for disease detecting, classifying andsurveying different types of authentic model is introduced byresearcher. For plant disease detection, classification andsurveying properly lots of innovative techniques areestablished by researcher and their work summery is includedin this section.Esker published a conference paper for detecting Stewart’sdisease on corn whose scientific name is “Pantoea stewartiisubsp” in 2006 [1]. They used three predictor model foridentify the stewartia corn disease and these three model nameare “Stevens”, “Stevens-boewe” and “Iowa state”. Amongthese three models “Stevens-boewe” finds the Stewart’sdisorder leaf blight phase. Umair Ayub published ainternational conference paper in Pakistan for finding cropdisorder dealing with Data Mining, 2018[2]. In this researchwork they mainly introduced losses which are faced byPakistani farmer and theses losses appear due to cropsdiseases which is occurred by the attack of insects. Foranalyzing the disorder properly they used several data miningmodel for instance Neural Network, Supporting VectorMachine, Decision Tree and K Nearest Neighbors ect. Jamesrethinks feature of Transgenic crops in 2002 and therequirement of maize over the world [3] and here introducedthat the corn approximate requirement id 852 million at 2020.The financial losses is caused by the heavy uses of pesticidesin corn is given by the Craig Osteen in the EconomicThreshold Concepts [4]. Ravi introduced a clear concept ofthe origination of peach, its biological action and Morphologyusing Medical Phytochemicals [5]. Here they mainly focusedon the use of peach fruits according to medicine and the use ofdifferent betterment of human being. Naeem identify work is licensed under a Creative Commons Attribution 4.0 International License.)292

Published by :http://www.ijert.orgInternational Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020manages fungal post-harvest pathogens of peach usingmorphological model [6]. They mainly characterize the fungalby morphological model and verified motive of postharvestrot of peach. In this work mainly focused on the detection ofplant diseases and provides appropriate cure instantly. Fordoing this work here used image processing and authentictechnologies like CNN so that the illiterate farmer can getimmediate result with high accuracy.III. PROPOSED METHODOLOGYIn our proposed model image processing method is usedfor the construction of system through which leaf disorder isdetected if any distorted picture is supplied with in very shorttime. As a result a farmer without sufficient sense diseasedetection knowledge, modern techniques and software can beeffortlessly applied this system. The dataset which is used asinput is mixed of healthy and distorted images and aftercompleting the action of input dataset the system outputprovides the affected and healthy leaves. A chart is introducedbelow as the proposed methodology. Figure 1 is thatflowchart.images 233, sort hole disease select 459 images, peach healthyimage number is 72, Alternaria blight disease select 80images, Anthracnose select 80 images, for downy mildew 94images 80 healthy images, powdery mildew select 80 images,Black rot uses 233 images. During training portion of theresearch work Foliar Fungal uses 410 images, select 953images as Gray leaf spot disease, Rust disease collect 929images, 929 maize healthy images, Short hole disease elected1838 images, peach fruits use 288 healthy images. The datasetwhich is collected before starting the research work isreshaped duo to match the picture size with each other and setthe pixel size as 265 256. In order to increase the imageproficiency the operation of image quality enhancement isreceived. The example of our selected dataset which occurbefore our research work starting is given below as Fig.2.Image AcquirementFundamental imageprocessingAction datasetImage ExpansionAnalytical analysisFig.1 Methodology Flow ChartA. Image AcquirementWhen one wants to start a research work his primaryresponsibility is to gather and process as many data as hecapable because in research work dataset contribute vitalaction. For obtaining perfect result and excellent exactnessand getting powerful research work need to collect adequatedata. During our research work we are able to collect around13000 corn, peach, grape, potato and strawberry leaf images.Major part of our dataset is collected from corn, peach, grape,potato and strawberry harvesting field and the remainder dataset has obtained from Google and public source of GitHub [7].For making the spontaneous process system we can take lotsof image format such as .gif, bmp, .jpg and so on. .B. Fundamental Image ProcessingForm total collected data we have elected around 10000valid data and the dataset separated into several folders. Thedataset which is selected as training and testing and its ratio is80 percent and 20 percent respectively. The testing sectionhave been selected 103 images for the disease of Foliar fungal,239 images for Gray leaf spot, Rust disease select 239images, 233 images for Common rust corn, maize healthyIJERTV9IS100123Fig.2. Accumulated DatasetC. System ArchitectureThe scheme is created using CNN Convolutional NeuralNetwork) multi-level model. The first convolutional layerinclude the ReLu activation function “1”, image input shapeis (256, 256, 3), 64 used as filter size, Kernel size (8 8),“SAME” Padding and the Strides is (1 1). The secondconvolutional layer display the equivalent shape of the firstlayer and the additional feature is Max Pool size (2 2) andstrides is (2 2).(1)In the third and forth convolutional layer ReLu activationfunction “1”, image input shape is (128, 128, 3), 32 used work is licensed under a Creative Commons Attribution 4.0 International License.)293

Published by :http://www.ijert.orgInternational Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020the filter size, Kernel size is (5 5), “SAME” Padding andthestrides is (1 1). The forth layer used Max Pool size (2 2)and strides is (2 2). In the fifth and sixth convolutional layerReLu activation function “1”, image input shape is (64, 64,3), 16 used as the filter size, Kernel size is (5 5), “SAME”Padding and the strides is (1 1). The sixth layer usedMax Pool size (2 2) and stride is (2 2). In the seventh andeighth convolutional layer ReLu activation function “1”,image input shape is (32, 32, 3), 8 used as the filter size,Kernel size is (3 3), “SAME” Padding and the strides is(1 1). The sixth layer used Max Pool size (2 2) and strideis (2 2). The flatten layer usage 512 units of the dense layerand among them 50 percent is dropped by the ReLuactivation function [10]. The utmost output tier used 5 unitswith softmax activation function “2”.(2)0.001 used as the learning rate used in our proposed model asthe optimization of ADAM [9].and weighting change. In our research work the categoricalcross entropy is used as the loss function “4”.(4)E. Image ExpansionThe image partitioning work is taken place in theprocedure of image expansion. The image expansion providessome motive and the main motive of the image expansion is: To represent the image through simplifying andalternating pattern. Changing the image shapes and angle forproducing superfluous data. Using maximum image rotation range is 40, thewidth and height shifting range is 0.2, rescalingvalue is 1/155, shearing and zooming range is0.2. During the expansion the horizontal flip actas True. For obtaining the greatest accuracy thenearest model is given in fig.4.Width shiftRotationFig.3. Proposed Convolutional Neural NetworkD. Optimization and Learning RateThe optimization algorithm is selected for confirming thesufficient variation of the result of deep learning andcomputer vision. The evaluation of various subsamples datais done in the Esoteric Adam paper. Various motive functionsare stabilizing in this paper. In the gradient steps theefficiency is shown enhancement pattern by the optimizationalgorithm [11]. Different types of application like NLP(Natural Language Processing) and computer vision showsperfect adaption with the optimization algorithm in themodern world. The optimization technique is efficient forindividual learning rate estimation dealing with numerousparameters from the 1st and 2nd moment of gradient. Themodel which is introduced here using 0.001 as the learningrate in the ADAM “3” optimization amidst.NormalHeight shiftOriginalZoomingRotated(3)The neural network and cross entropy function providesacceptable output in the modern work of classification andprediction and this result is more grantable than MSE (meansquare error). Usually, the training is not stalled out due todisable of getting sufficient minor using Cross-entropy errorIJERTV9IS100123Width shiftHeight ShiftZoomFig.4. Example of Image work is licensed under a Creative Commons Attribution 4.0 International License.)294

Published by : International Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020The angle of the image is rotated counterclockwise which is controlled by the sheer rangeand allow our images to be sheared.The image is “Rescale” by multiplying theimage data with numeric value at the initial stateof the image processing. The coefficient range ofthe image is 0-255 and it is known as RGBimage. But the range of the image is very high inour proposed model. As a result the range of ourtarget value is one and zero and that valueacquired by scaling the images with 1/255.F. Training the ModelNumerous validation and training dataset is used fortraining the model with the batch size 30. At the time ofoperation, the validation and reduction accuracy rate issupervised using learning rate reduction method. When 30epochs is completed the supervision is worked manuallyamong the validation accuracy and reduced learning rate.Then our model process 15-20 epochs and in a certain timeset causal learning rate.Fig.7. Layer visuallization using kernel size (3,3)G. Layer VisualizationThe gradual change of image is symbolized by layervisualization. The visual change of image in multiple layers isgiven below.Fig.8. Layer visuallization using kernel size (2,2)IV. RESULT AND DISCUSSIONIn our research work, the proposed model is providedexpected output after completing the training, testing andvalidation using various dataset. The detail description aboutthis model output is given below.Fig.5. Layer visuallization using 3 3 matrixFig.6. Final layer visuallization using 2 2 matrix formatIJERTV9IS100123A. Analytical AnalysisThe training and validation accuracy is obtained throughthis model 77.99% and 34.17% respectively. With time, aftereach run the model experienced trained and the exactness ofthe result improved as well. After completing 10 runs thetraining and validation accuracy become 89.56% and 61.91%respectively and the learning rate decreased at .0005. Whenthe successful run is counted 30 the training and validationaccuracy reached to 93.8% and 95.8% respectively and thelearning rate decreased at 3.124e-05. After completing finalrun the height accuracy is obtained and the value for trainingand validation accuracy is 94.29% and 96.28% respectively.B. Accuracy GraphThe term over fitting is used for describing and referring amodel and this appear during capturing noise of the data. Sothe indication is that the perfect fitting of model oralgorithmic data occur over fitting. The term under fitting isintroduced during showing the lacking of capturinginsufficient underlying data. The model which we introducedhere shows absence of over or under fitting. The graph fortraining and validation accuracy and loss of our model isshown in Fig work is licensed under a Creative Commons Attribution 4.0 International License.)295

Published by :http://www.ijert.orgInternational Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020TABLE I.CLASSIFICATION REPORTDiseaseprecisionrecallF1-scoreAlternaria blight0.781.000.76Anthracnose0.970.860.91Downy mildew0.920.970.94Healthy leaf0.800.890.92Healthy flower0.880.840.84Unhealthy flower0.940.950.96Gray leaf spot0.970.940.96Common Rust0.990.990.99Healthy corn0.990.990.99Bacterial spot0.980.980.98Healthy peach0.970.980.97Micro avg.0.920.920.92Macro avg.0.920.920.92Weighted 20412041204D. Result Analysis of Different ModelSome research papers are analyzed by our self which arerelated to our research work. We compared out model withother researchers model and found that the model which weproposed provide best result shown in Table 2Fig:-9 Training and validation accuracy and loss graph.C. Confusion MatrixThe performance of the model is shown by the specificerror table known as confusion or error matrix. Among total67 images true image is 67 and false image is 0 only forAlternaria blight, for Anthracnose true images is 41 and falseimage is 45 among total approximate image 86, The Dwnymilddew shows 91 true and 3 false image among 94, ForPowdery milddew consider total 80 images where 44 imagesare true and the remain 36 is false. In Healthy, among 70image 59 detected as true and 11 is false. Finally for Blackrot, among total 80 images 58 shows true and 22 is false. Thevalues introduced in diagonal position of the confusionmatrix are bigger than others. The values in diagonalposition used (4 4) shape is showing their best performancecomparing to the other position and this part maximizing thedata and that’s why its color is deep blue. Fig.10. Showsconfusion matrix and Table 1 shows classification report ofour model.TABLE II.WorkACCURACY COMPARISION BETWEEN MODELSAccuracy (%)workAccuracy(%)Sharada et al.85.53s.phadikar et al.79.50[12][15]Prem et al. [14]89.93Jyoti and tanuja93.00[16]Proposed94.29Naik Durgesh et94.00modelal. [17]E. Result Analysis and Variatioon AccuracyThis proposed model worked well for the separate datasetof corn and peach, shown in (Table 3).TABLE asetSEPERATE RESULT RATION IN OUR . CONCLUSIONThis work provides an authentic notion for detecting theattacked leaf (‘Grape’, ‘Potato’ and ‘Strawberry’) and thefarmer who works for produce these fruits gets remedy so thatthey can enhance the production in agricultural industry.Specialist who works in agriculture department accepts quickdisease detection process by image processing technique as aresult Image Processing technology touch it’s milestonewithin very short time. The transited portion of leaf easilysegments and analyzes using CNN model and this modelprovides best possible result instantly. As a result the farmerwho detects plant disease manually can save their time anddiminish suspicion on possibilities of wrong detection. Ourfuture goal is to develop an open multimedia system and makea software which automatically detect plant disease andprovide their solution.Fig.10. Confusion work is licensed under a Creative Commons Attribution 4.0 International License.)296

Published by :http://www.ijert.orgInternational Journal of Engineering Research & Technology (IJERT)ISSN: 2278-0181Vol. 9 Issue 10, October-2020REFERENCES[1]Esker, P. D., Harri, J., Dixon, P. M., And Nutter, F. W., Jr. 2006.Comparison Of Models For Forecasting Of Stewart’s Disease Of CornIn Iowa. Plant Dis. 90:1353-1357.[2] Umair Ayub, Syed Atif Moqurrab: Predicting Crop Diseases UsingData Mining Approaches: Classification. 2018 1st InternationalConference On Power, Energy And Smart Grid (Icpesg). Doi:10.1109/Icpesg.2018.8384523.[3] James, C. 2003. Global Review Of Commercialized Transgenic Crops:2002 Feature: Bt Maize. Isaaa Briefs No. 29. Isaaa: Ithaca, Ny.[4] Craig Osteen, A.W. Johnson, And Clyde C. Dowler : Applying TheEconomic Threshold Concept To Control Lesion Nematodes On Corn,Natural Resource Economics Division, Economic Research Service,U.S. Department Of Agriculture.Technical Bulletin No. 1670.[5] Ravi Kant, Rishi Kumar Shukla, Abha Shukla: A Review On Peach(Prunus Persica): An Asset Of Medicinal Phytochemicals. InternationalJournal For Research In Applied Science & Engineering Technology(Ijraset) Issn: 23219653; Ic Value: 45.98; Sj Impact Factor :6.887Volume 6 Issue I, January 2018[6] Naeem, M., Irshad, G., Naz, F., Noorin, S., Aslam, F., & Rafay, A.(2018). Morphological Identification And Management Of FungalPost-Harvest Pathogens Of Peach (Prunus Persica L). World Journal OfBiology And Biotechnology, 3(1), 183-185[7] Tree/Master/Dataset. [Accessed: 23-April 2019].[8] ,Salakhutdinov,R.:Dropout:Asimple Way To Prevent Neural Networks FromOverfitting. J. Mach. Learn. Res. 15 1929–1958 (2014)[9] uralnetw orksInclassification (2017). Arxiv:1702.05659[10] Goutum Kambale1, Dr.Nitin Bilgi : A Survey Paper On Crop DiseaseIdentification And Classification Using Pattern Recognition AndDigital Image Processing Techniques. May (2017).IJERTV9IS100123[11] M. Akila And P. Deepan : Detection And Classificationof Plant LeafDiseases By Using Deep Learning Algorithm.(2018). InternationalJournal Of Engineering Research & Technology (Ijert) Issn: 2278-0181Published By, Www.Ijert.Org Iconnect - 2k18 Conference Proceedings[12] G. Prem Rishi Kranth, M. Hema Lalitha, Laharika Basava, AnjaliMathur : Plant Disease Prediction Using Machine LearningAlgorithms. November (2018). International Journal Of ComputerApplications (0975 – 8887) Volume 182 – No. 25.[13] S. Phadikar And Et Al: Baye’s And Svm Classifier , Mean FiiteringTechnique And Otsu’s Algorithm.[14] Jyoti And Tanuja : Cotton Plant Leaf Diseases Identification UsingSupport Vector Machine. International Journal Of Recent ScientificResearch Vol. 8, Issue, 12, Pp. 22395-22398, December, 2017[15] W.-K. Chen, Linear Networks and Systems (Book style).Belmont,CA: Wadsworth, 1993, pp. 123–135AUTHOR PROFILEMd. Tariqul Islam received the B.Sc.Engineering (ECE) degree form KhulnaUniversity of Engineering and Technology(KUET), Bangladesh, June 2017. He iscurrently working as a Lecturer inDepartment of Computer Science andEngineering(CSE)inBangladeshUniversity, Dhaka, Bangladesh where hejoined in January, 2018. His ma

maize act as a good source of minerals, dietary fiber and vitamins. There exist different types of corn for instance pop corn, dent corn, flour corn, sweat corn and flint corn. Spring season consider as the best time period for maize plantation and the corn is unable to tolerate coolness. The plant grows rapidly with the moisture soil.

Related Documents:

CNN R-CNN: Regions with CNN features Figure 1: Object detection system overview. Our system (1) takes an input image, (2) extracts around 2000 bottom-up region proposals, (3) computes features for each proposal using a large convolutional neural network (CNN), and then (4) classifies each region using class-specific linear SVMs. R-CNN .

Fast R-CNN [2] enables end-to-end detector training on shared convolutional features and shows compelling accuracy and speed. 3 FASTER R-CNN Our object detection system, called Faster R-CNN, is composed of two modules. The first module is a deep fully convolutional network that proposes regions, and the second module is the Fast R-CNN detector [2]

A deep CNN is learned to jointly optimize pedestrian detection and other semantic tasks to im-prove pedestrian detection performance [32]. In [5,36,20,40,33,4], Fast R-CNN [16] or Faster R-CNN [27] is adapted for pedestrian detection. In this paper, we explore how to learn a deep CNN to improve performance for detecting partially occluded .

Fast R-CNN a. Architecture b. Results & Future Work Agenda 42. Fast R-CNN Fast test-time, like SPP-net One network, trained in one stage Higher mean average precision than slow R-CNN and SPP-net 43. Adapted from Fast R-CNN [R. Girshick (2015)] 44.

Modern Object Detection Architecture (as of 2017) Multiple important works around 2014-2017 which built the basis of modern object detection architecture R-CNN Fast R-CNN Faster R-CNN SSD YOLO (v2, v3) FPN Fully convolutional . Let's dissect the modern (2017) object detection architecture! Detectron 9

High Brow Hickory Smart Little Kitty Smart Lil Highbrow Lena Joe Peppy 1992 Consigned by CNN Quarter Horses CNN Highbrow Lady 2006 Bay Mare CNN Highbrow Lady 4902100 NOTES: CNN Highbrow Lady is a smart, fancy, Highbrow filly out of a powerful female line. She is well broke.

fast-rcnn. 2. Fast R-CNN architecture and training Fig. 1 illustrates the Fast R-CNN architecture. A Fast R-CNN network takes as input an entire image and a set of object proposals. The network first processes the whole image with several convolutional (conv) and max pooling

Jia-Bin Huang, Virginia Tech. Today's class Overview Convolutional Neural Network (CNN) Understanding and Visualizing CNN Training CNN. Image Categorization: Training phase Training . CNN as a Similarity Measure for Matching FaceNet [Schroff et al. 2015] Stereo matching [Zbontar and LeCun CVPR 2015]