Banana Plant Disease Classification PDF

Hindawi Computational Intelligence and Neuroscience Volume 2022, Article ID 9153699, 13 pages https://doi.org/10.1155/2022/9153699 Research Article Banana Plant Disease Classification Using Hybrid Convolutional Neural Network K. Lakshmi Narayanan ,1 R. Santhana Krishnan ,2 Y. Harold Robi...

Hindawi Computational Intelligence and Neuroscience Volume 2022, Article ID 9153699, 13 pages https://doi.org/10.1155/2022/9153699 Research Article Banana Plant Disease Classification Using Hybrid Convolutional Neural Network K. Lakshmi Narayanan ,1 R. Santhana Krishnan ,2 Y. Harold Robinson ,3 E. Golden Julie ,4 S. Vimal ,5 V. Saravanan ,6 and M. Kaliappan5 1 Department of Electronics and Communication Engineering, Francis Xavier Engineering College, Tirunelveli, India 2 Department of Electronics and Communication Engineering, SCAD College of Engineering and Technology, Tirunelveli, India 3 School of Information Technology and Engineering, Vellore Institute of Technology, Vellore, India 4 Department of Computer Science and Engineering, Anna University Regional Campus, Tirunelveli, India 5 Department of Artificial Intelligence and Data Science, Ramco Institute of Technology, Rajapalayam, India 6 Department of Computer Science, College of Engineering and Technology, Dambi Dollo University, Dembidolo, Ethiopia Correspondence should be addressed to V. Saravanan; [email protected] Received 27 November 2021; Revised 20 January 2022; Accepted 30 January 2022; Published 23 February 2022 Academic Editor: Alexander Hošovský Copyright © 2022 K. Lakshmi Narayanan et al. 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. Banana cultivation is one of the main agricultural elements in India, while the common problem of cultivation is that the crop has been influenced by several diseases, while the pest indications have been needed for discovering the infections initially for avoiding the financial loss to the farmers. This problem will affect the entire banana productivity and directly affects the economy of the country. A hybrid convolution neural network (CNN) enabled banana disease detection, and the classification is proposed to overcome these issues guide the farmers through enabling fertilizers that have to be utilized for avoiding the disease in the initial stages, and the proposed technique shows 99% of accuracy that is compared with the related deep learning techniques. 1. Introduction 20% of the global banana production. The United States is the leading importer of bananas of about 18% of the global Agriculture is the crucial resource of food for mankind, and imports. The impact of the banana tree getting infected due it is one of the important factors that decide the economy of to disease and due to other climatic changes will cause even a country. Agriculture is considered the main source of 100% loss in the overall countries’ banana production and income for most developing countries. One of the important export. Generally, bananas are affected by four major that parts of the global agro-business is the banana cultivation or are black Sigatoka, fusarium wilt colloquially called Panama banana industry because bananas are rich in minerals such wilt, Xanthomonas wilt, and bunchy top virus. The details of as calcium, manganese, potassium, magnesium, and iron. As the various commonly found diseases along with the disease this particular crop is having these many vitamins, they are symptoms, appearance, and effects of the disease when it is consumed by people all over the world as banana is con- present as an infection in the banana leaf is described below. sidered an instant energy booster. As per the statistics from Wikipedia, about 15% of the global banana productions are exported to western countries for consumption. As per the 1.1. Banana Leaf Disease. The banana crops are all affected production and export statistics of bananas, about 25.7% of by various diseases. The symptoms are visible in leaf, stem, the global banana production is from India, and other major flower, fruit, roots, and suckers. The major diseases that producers of bananas are the Philippines, Ecuador, Indo- affect the leaf are Xanthomonas wilt, fusarium wilt, black nesia, and Brazil giving a combined contribution of about and yellow Sigatoka. 2 Computational Intelligence and Neuroscience 1.1.1. Xanthomonas Wilt. Banana Xanthomonas wilt become darker and turn into depression, which is dem- (BXW) is a bacterial infection caused by Xanthomonas onstrated in Figure 4. campestris. Among the numerous diseases infecting banana, This may decrease the yield by 30 to 50% depending the damage caused by BXW has been huge. The production upon the severity of the disease infection. The common of bananas is decreased by 30–52% due to BXW. The in- remedies for this disease are by spraying 3 times with fected plant appears in pale yellow-orange color and in the carbendazim or propicanozole or mancozeb and tempol in later stage it becomes dark brown color. At last, it leads to proper proportion at 10–15 days period, as the disease starts death, and this may cause 100% yield loss if it is not managed from the original look of leaf specks. Therefore, the detection well. The remedy for clearing this disease is by field of the pest and disease at an earlier stage is very important. sanitation and removal of affected plant parts or by spraying The traditional methods used for the identification of pests streptocycline 200 ppm after the first visual symptom for and diseases are limited by the lack of human knowledge every 10 days. Further by spraying 0.3% copper oxychloride, based on agriculture , so deep learning-based hybrid it checks the further spread and is demonstrated in Figure 1. CNN and FSVM are developed for detecting and classifying the above mentioned common disease found in the banana leaf. This helps in detecting the disease and classifies its type 1.1.2. Fusarium Wilt. It is usually identified as Panama in its early stage. Through this banana, production can be infection, which is a deadly fungal infection originated from effectively managed. the soil-borne fungus Fusarium oxysporum. This disease can The rest of the paper is ordered in such a way that related cause 100% yield loss if it is not identified and managed in works in the field of plant disease detection that is carried out the early stage. The fungus penetrates the plant into the in Section 2, materials and methods with the proposed root and settles the xylem vessels, thus jamming the water method of detecting the banana leaf detection is proposed in flow. It appears like a pale yellow color in the early stage and Section 3, and Section 4 discusses the results and discussion in the later stage, it looks dark in color. It is irregular in with a comparison of the proposed system with the existing shape, pale margin on new leaves, leaf blades are distorted. system, and finally conclusion and future discussion in Fusarium wilt is a disease that the fungus assaults the Section 5. vascular tissue within the root discoloration. The fruit does The main objectives of the proposed model are not show any symptoms. The first sign of the disease is wilting and produces the yellow color of the older leaf at the (1) The main objective of this work is to apply deep margin. The tear of the base is the main effect that the learning methodology for the detection of evident infected sucker does not show any symptoms until 4 months. banana disease in the banana plant Suggested remedies for this disease infection are applying (2) To classify the type of disease with high accuracy 2.5 kg/ha Pseudomonas fluorescent in the farmyard along (3) By creating a database of insecticides for respective with regular manure and it is illustrated in Figure 2. pests and diseases, it will provide a remedy for the disease that is detected 1.1.3. Bunchy Top Virus. It is a plant pathogenic virus an- (4) As CNN and SVM are the best classifiers available cestor nanoviridae known for contaminating banana plants. with a hybrid model of combining both the algo- It is a viral infection caused by a single-stranded DNA virus rithms to produce accelerating performance known as the banana bunchy top virus. This disease is af- fected in a tropical region and transmitted from plant to 2. Related Works plant. Symptoms for this infection commonly occur in old plants in which the latest leaves are narrow yellow than The Banana leaf disease detection and classification is an normal , and the banana bunchy top virus is illustrated in issue for a very long time many kinds of research were done Figure 3. The common remedies for this disease are by on this field which gave truthful results details of a few such injecting 4 ml of fernoxone solution along with 400 ml of findings were given below. water or by injecting 4 ml of monocrotophos on a 1 : 4 ratio An artificial intelligence-based banana disease and pest at 45 days interval till flowering from the 3rd month. And detection was proposed in , here the algorithm used is also by spraying either 1 ml of phosphamidon or 2 ml of deep convolutional neural network for disease detection, methyldemeton or 1 ml of monocrotophos along with water here the author has collected data sets for about 8 different can avoid the disease spread. diseases in banana, a sum of 30,000 images were used as a data set, and the proposed system produces a result of 90% accuracy. Machine learning algorithms were developed for 1.1.4. Black Sigatoka. It is called a black leaf streak caused by detecting and classification of plant diseases , reviews mycosphaerella Fijians is that the Sigatoka infection mul- various machine learning and deep learning algorithms for tifaceted is a bunch of intimately correlated fungi. Leaves detection and classification of plant diseases, and this also with huge infectious lesions will begin to collapse, and it identifies some research gaps for detecting the disease in the interrupts performing photosynthesis, which leads to the plants even before the symptoms are visualized. death of plants. In the early stage of infection, the lesions A deep leaning-based banana leaf disease classification is have a rusty brown look. After further development, they proposed , here they have used LeNet architecture as a Computational Intelligence and Neuroscience 3 Figure 1: Banana Xanthomonas wilt. Figure 2: Banana fusarium wilt. Figure 3: Banana bunchy top virus. 4 Computational Intelligence and Neuroscience technique based on tomato disease identification have been implemented with two different techniques as Faster R–CNN, which is used for identifying the type of tomato disease and Mask R-CNN, Which is used to find and seg- ment the location and shape of the infected areas, and the results show that the proposed method gives an accuracy of 90% in detecting the disease and 99% in identifying the shape of the infection. Automatic detection and classification of diseases in rice crop are proposed in , and here they use an artificial neural network-based technique to identify the disease, they have done research on many diseases that affects the rice crop and measured the accuracy of classifier for each type of disease, and finally they have compared the ANN technique with other leading classifiers for detecting the same diseases and their accuracy were measured and concluded that ANN Figure 4: Banana with black Sigatoka. gives a satisfactory result when compared with the other classification algorithms. A novel rice blast recognition convolutional neural network for classifying the data. The technique related to CNN is proposed in. This method results of this research demonstrate its effectiveness in was tested with various combinations such as CNN only, various conditions of the images such as complex back- CNN with SVM, LBPH with SVM, and Haar-WT with SVM ground, different size, and orientation. This method gets and their accuracy is compared, and it shows that the CNN stabilized in 25 iterations and achieves a good accuracy at the with SVM gives an accelerating accuracy of about 96% AUC final iteration. A machine learning-based approach for curve that shows 0.99. The drawbacks of the existing works are detection of banana disease detection in the early stage using The main drawback of the existing system is that most the SVM classifier is proposed. Here the images used are of the works follow image processing techniques that close-range hyperspectral remote sensing images. The results involve complex image segmentation steps, which is a of the classifiers are evaluated by overall accuracy, and very time-consuming process average accuracy here is the accuracy using spectral and morphological information, which is about 96% in early Many diseases will not show their symptoms by having detection, 90% in mid detection, and 92% in late detection. a clear edge, and they may merge with the healthy parts An image processing-based banana leaf disease detection of the leaf, which cannot be detected using the existing is proposed in , here the images are first acquired and the techniques and requires a powerful classification RGB model is converted into an HSI color model and then algorithm preprocessed, then the image is segmented using the Certain techniques that are available will be suitable for thresholding method, and the histogram equalization is a particular type of crop or limited to certain crops only found for HSI image. Then the classification is compared Methods such as ANN, KNN, PCA, and other image with three different classifiers such as they are back- processing techniques are lagging in accuracy and propagation neural networks, SVM and principle compo- mostly consume more time to classify the disease nent analysis (PCA). An artificial neural network-based In most cases, the infections in the banana plant will banana leaf disease detection and classification of the disease occur in various parts of the plant, but most of the is proposed in , here the image is acquired and pre- research concentrates only on the leaf processed initially and then the color and HOT (Histogram of Template) feature are extracted, then the data set is trained using the artificial neural network, and then the grading is 3. Materials and Methods done on for the query images based on the total percentage of the affected area. And finally, the image is classified by its 3.1. Dataset Collection. The dataset for this proposed re- disease type. search consists of around 3500 images of banana plants, both Not only for banana leaf but also more research is going infected and healthy parts of banana plants were collected on for detecting and classifying the disease in most con- from various fields located in south India, especially the sumable crops like paddy, maize, apple, cheery, and other southern part of Tamil Nadu from the districts of Madurai, common plants. Some of those findings are given below. Dindigul, Virudhunagar, Tirunelveli, Tuticorin, Nagercoil, Classification of apple plant and cherry plant diseases using Kanyakumari, and minor parts of Tamil Nadu Kerala improved convolutional neural networks is proposed in. border. The data sets were collected in a balanced number of The improvement in the CNN is based on merging a images for 4 basic diseases that can influence the produc- framework of inception functionality squeeze, excitation tivity of banana production. The images were collected in functionality, and a global pooling layer. The results of this various resolutions captured using mobile phones with good method show an accuracy of about 91.7% on the test data set. resolutions, captured using VGA cameras in mobile, Digital A deep convolutional neural networks and object detection cameras, and DSLR cameras. Computational Intelligence and Neuroscience 5 3.2. Image Preprocessing. Image preprocessing on the where E(δ) is the highest estimation parameter, and the collected image is performed to enhance the image quality vector values are computed using for performing the further steps. This process does not change the default information in the image; it performs n image resizing and performs some useful filtering pro- ωn 􏽘 ωni. (4) cesses to detect the disease information in the banana leaf. i 1 The image captured and stored as a data set is of different The next operation is pooling that reduces dimensions of resolutions so it must be converted into a standard fixed the feature maps, and this helps in reducing the computa- resolution size using image resizing. Then by using image tions to be performed in the network. So further operations filtering techniques such as median filter other noises will be performed on the précised positioned features present in the images are removed. Focus issues and other generated by the convolutional layer. There are multiple unwanted portions in the image while capturing are re- types of pooling methods available in this context of the Max stored in this process. Another filtering process such as low pooling method so that the output of the Max-pooling layer pass filtering that helps in reducing the amplitude of high- will contain the most important features of the earlier feature frequency components in the image and keeps the low- map. This pooling-based reduction is done without losing frequency information as it is. The high-pass filter does the the features or patterns. vice versa and the proposed architecture is illustrated in After completing a sequence of convolution layers, ReLU Figures 5 and 6. activation, and pooling operations, the next step is to per- form the flattening operation that converts the 2D matrix of features into a vector of features, which is then fed into a 3.3. Feature Extraction Using CNN. The main objective of the classifier model. The main objective of the fully connected CNN is to recover the high-level features of the banana leaf process is to feed the flattened vectors into the classifier, and infections in the image. To perform this, CNN architecture the 2D Max pooling operations in Keras are demonstrated in was built with multiple overlying convolutional layers. Figure 8. Normally, a convolutional layer will contain convolution In this proposed work, train and test the images in and activation functions that are non-linear in our case, and various proportions they are 90%, 70%, 50%, 30%, and 20% the utilization of ReLU and pooling process have been as training images, and the remaining images are treated as enabled. The input for the CNN will be training and testing the test images. With these proportions, compare the per- images of dimensions of 80 × 120 in height and width, which formance of classification on training accuracy and testing is demonstrated in Figure 7. accuracy. The proposed CNN architecture is motivated by LeNet- 5. The convolutional layers are called C1, C2 which contains 4, 8, and 16 filters with dimension 5 × 5, 3 × 3 and 2 × 2, respectively. Let vl denotes the output of the convolutional 3.4. Proposed Fusion SVM Classifier. SVM is a supervised layers, and it is expressed in machine learning algorithm and is one of the most popular approaches for image classifications. SVM is a binary X Y classification algorithm that classifies the images into only vl f ⎛ ⎝al + 􏽘 􏽘 gl hl−1 ⎞ ⎠ x,y x,y. (1) two distinct classes, infected or not infected. But the real- x y world problem is to identify the exact infection type, which Here, involves the classifier to classify the image into more than two classes or multiple classes. There are many types of X, Y represents size of the filters (height and width) methods that can be followed to use SVM for multiclass al represents bias of the convolutional layer classification such as One versus All, One versus One, hl−1 represents the output of the preceding convolu- and All versus All. tional layer In this proposed work, the SVM is used in two dif- ferent phases P1 and P2. In the first phase P1, the feature gl represents the weight of the convolution layer extracted from the CNN model is fed as an input to a The non-linear activation function of ReLU (f(v)) is binary SVM model. Here the model classifies the leaf computed in image and other parts of the banana plant available in the data set as infected on a healthy leaf. If the test image of the v, v > 0, f(v) ReLU(v) f(v) 􏼨 (2) leaf given as an input is a healthy leaf, then the process 0, v ≤ 0. terminates. In the second phase, P2 construct a multiclass support vector machine to predict the disease type. Given The parameter δ could be computed using the highest a new image, our proposed model determines whether the estimation with the particular training set, and it is com- banana plant is infected or not. If it is an infected image, puted in then it predicts its class out of 4 available disease classes. N The proposed classification method is described in E(δ) 􏽙 f(v)δn , (3) Figure 9. n 1 6 Computational Intelligence and Neuroscience Discard Healthy Images Infected Image BXM Classification Results BFM Feature Extraction using Image Preprocessing Binary SVM Multiclass SVM CNN BBTV Vectors BBS Figure 5: Block diagram of the proposed method. Image Resizing Full commected Image Max-Pooling RELU Pre-processed Feature Extraction Hyper plane 1 By Class 1 Class 2 Convolutional Neural Networks XANTHOMONAS WILT FUSARIUM WILT Hyper plane 2 Vectors BUNCHY TOP VIRUS Class 3 Class 4 BLACK SIGATOKA Classification By Multiclass SVM Figure 6: Overview of the proposed method. 16@56×36 16@28×16 8@114×74 4@116×76 8@57×37 1×128 Convolution Convolution Max-Pool Convolution Max-Pool Figure 7: Architecture of the convolutional neural networks for the proposed system. Computational Intelligence and Neuroscience 7 2 2 7 3 Max Pooling 9 5 6 1 9 7 8 7 3 5 Filter (2×2) 8 6 3 2 1 6 Stride (2, 2) Figure 8: 2D Max pooling operations in Keras. Infected Fully connected Features Healthy Vectors from the Convolutional Neural Networks Binary SVM Hyper plane 1 BXW BFW Is Banana Yes Leaf Infected? Hyper plane 2 BBTV BBS No Multi Class SVM STOP Figure 9: Structure of proposed fusion SVM classifier. (1) Input: features from the CNN model (2) Procedure Support Vector Machine (3) Binary classifier P1 classifies the input image as infected or healthy. Training and Testing (4) If Classifier Result of P1 Healthy (5) Process terminated. (6) Evaluate The Binary Classification Result (7) If Classifier Result of P1 Infected (8) Develop multiclass classifiers P2 (9) End Procedure (10) Output: Disease Classification (11) Evaluation of the Classification Result ALGORITHM 1: Proposed fusion SVM classifier. 4. Results and Discussions based SVM testing operation, the first-level P1 binary SVM classifies the image as infected leaf or a healthy leaf. If the The testing of the banana tree disease detection and clas- given test image is a healthy image, the process exits in the sification was performed using PYTHON programming phase P1, and the classifier performs the training process language in Jupyter notebook environment on an I5 pro- when a new image is given as a query image for performing cessor with 32 GB of RAM equipped with 6 GB AMD GPUs itself. If the test image is infected, then the features are from NVIDIA. To demonstrate the proposed method of applied to the second phase P2, which is a multiclass image classification, the experimentation has been con- classification. Here the infected banana plant is classified as ducted with 5 categories of images taken. The proposed either one of the four classes, namely banana Xanthomonas system takes the image data and extracts the image features wilt (BXW), banana Fusarium wilt (BFW), banana bunchy using the proposed CNN whose architecture is inspired by top virus (BBTV), and banana black Sigatoka (BBS). Then LeNet-5. Then with the extracted features of the fusion- the performance of the proposed classifier is measured 8 Computational Intelligence and Neuroscience Figure 10: Image classified as BXW showing the accuracy and elapsed time. Figure 11: Stem image classified as BXW showing the accuracy and elapsed time. according to the accuracy in percentage and elapsed time fusarium, the treatment is to spray Propiconazole, and if (ET) in seconds. Simultaneously, a confusion matrix is the predicted disease is bunchy top, Fernoxone is the constructed to calculate the classification precision, recall, possible treatment. Figure 11 shows the query image and F1 score. classified as black Sigatoka with an accuracy of 99% and Figure 10 shows the output of the proposed system elapsed time of 148.51 seconds. Figure 12 illustrates the with a classification accuracy of 99% and elapsed time of image classified as BBS showing the accuracy and elapsed 19.53 seconds. Furthermore, this output suggests the time. possible treatments to be given for the detected disease To measure the performance of the model proposed a such as if the disease is black Sigatoka, the treatment is confusion matrix is constructed. The confusion matrix is a Carbedazime; if the predicted disease is Xanthomonas, the tabular outline representing the efficiency of the proposed treatment is to spray Mancozeb; if the predicted disease is model and how it performs. Figure 13 shows the confusion Computational Intelligence and Neuroscience 9 Figure 12: Image classified as BBS showing the accuracy and elapsed time. A 0.96 0.01 0.01 0.02 0.8 B 0.03 0.93 0.02 0.02 0.6 0.4 C 0.04 0.13 0.82 0.02 0.2 D 0.03 0.05 0.02 0.9 A B C D Figure 13: Confusion matrix. matrix of the proposed model, there are 4 elements used in TN + TP the confusion matrix they are. accuracy × 100, Total True positive (TP): proposed model predicts true, while TP actually it is true precision , TP + FP True negative (TN): proposed model predicts false, (5) while actually it is false TP recall(or)sensitivity , False positive (FP): proposed model predicts true, while TP + FN actually it is false 2 ∗ precision ∗ recall False negative (FN): proposed model predicts false, F1 score. precision + recall while actually it is true Table 1 shows the performance of the proposed model Accuracy tells the overall correct prediction, the recall evaluated with different test images. The metrics used to is the measure of positives that are misclassified as neg- estimate the performance of the proposed model are pre- ative, precision is the measure of negatives that are cision, recall, accuracy, and F1 score. The following are the misclassified as positives, F1 sore is the harmonic mean of equations for calculating the performance metrics. precision and recall. Figure 14 shows the accuracy levels of 10 Computational Intelligence and Neuroscience Table 1: Performance evaluation of the proposed classification. Class F1 score Accuracy (%) Recall Precision BXW 0.96 98 0.98 0.94 BFW 1 99.90 1 1 BBTV 0.99 99.60 0.98 1 BBS 0.95 97.70 0.94 0.97 101 100 100 99 Accuracy (%) 99 98 98 97 97 BXW BFW BBTV BBS Classes Figure 14: Accuracy of the classification. 1.01 1 0.99 0.98 0.97 0.96 0.95 0.94 0.93 0.92 0.91 BXW BFW BBTV BBS Precision Recall F1 Score Figure 15: Performance evaluation. different classes predicted by the model, and Figure 15 comparison was made based on the literature survey shows the performance of other metrics. comparing various classifiers such as CNN, SVM, random Figure 16 shows the ROC curves plotted between the forest, SVM + PCA, ANN, CNN + RF, and SVM + Haar-WT. true-positive rate and the false-positive rate for the proposed Comparison results show that CNN + fusion SVM-based method, and Figure 17 shows the comparison of the pro- classifier accelerates performance for classifying multiple posed model with other classifiers based on its accuracy. This classes. Computational Intelligence and Neuroscience 11 ROC curve of the Banana Disease Classifier 1.2 1 True Positive Rate 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 1.2 False Positive Rate Figure 16: Receiver operating characteristic (ROC) curves for the proposed method. Proposed CNN+FSVM SVM+Haar-WT CNN+RF Classifiers SVM+PCA ANN Random Forest SVM CNN 84 86 88 90 92 94 96 98 100 Accuracy Figure 17: Comparison of various classifiers with the proposed method. Table 2: List of abbreviations. Abbreviation Acronyms CNN Convolution neural network FSVM Fusion support vector machine BXW Banana Xanthomonas wilt BFW Banana fusarium wilt BBTV Banana bunchy top virus BBS Banana black Sigatoka RGB Red green blue HSI Hue, saturation, and intensity PCA Principle component analysis HOT Histogram of template R–CNN Region-based convolution neural network ANN Artificial neural network LBPH Local binary pattern histogram Haar-WT Haar-wavelet transform AUC Area under curve KNN K-nearest neighbors VGA Video graphics array DSLR Digital single-lens reflex ReLU Rectified linear unit 12 Computational Intelligence and Neuroscience Table 2: Continued. Abbreviation Acronyms AMD Advanced microdevices GPU Graphics processing unit ET Elapsed time TP True positive TN True negative FP False positive FN False negative ROC Receiver operating characteristic RF Random forest 5. Conclusion and Future Scope Agronomy for Sustainable Development, vol. 39, no. 2, p. 22, 2019. Agriculture is very significant in producing food for R. C. Ploetz, “Management of Fusarium wilt of banana: a humanity and plays an important role in the growth of the review with special reference to tropical race 4,” Crop Pro- country’s economy. So the proper management of the tection, vol. 73, pp. 7–15, 2015. agricultural products is very important. Banana is a very J. Qazi, “Banana bunchy top virus and the bunchy top dis- important and vital crop produced all over the world, and ease,” Journal of General Plant Pathology, vol. 82, no. 1, pp. 2–11, 2015. it is necessary to prevent bananas from getting infected T. Yonow, J. Ramirez-Villegas, C. Abadie, R. E. Darnell, due to harmful diseases such as banana Xanthomonas wilt N. Ota, and D. J. Kriticos, “Black Sigatoka in bananas: Eco- (BXW), banana Fusarium wilt (BFW), banana bunchy top climatic suitability and disease pressure assessments,” PLoS virus (BBTV) and banana black Sigatoka (BBS). In this One, vol. 14, no. 8, Article ID e0220601, 2019. research, an integrated deep learning-based solution has M. H. Saleem, J. Potgieter, and K. Mahmood Arif, “Plant been proposed for detecting and classifying the banana disease detection and classification by deep learning,” Plants, disease by investigating not only the banana leaf but also vol. 8, no. 11, p. 468, 2019. with other parts of the plant with the help of CNN and an M. A. Lopez, J. M. Lombardo, M. López, D. Álvarez, FSVM, which is a combination of binary and multiclass S. Velasco, and S. Terrón, “Traceable Ecosystem and strategic SVM. The proposed method shows an accelerating per- framework for the creation of an integrated pest management formance of an overall accuracy of 99%, with an excellent system for intensive farming,” International Journal Of In- precision-recall and F1-score, which shows that the teractive Multimedia And Artificial Intelligence, vol. 6, no. 3, proposed method is best suitable for detecting the disease pp. 47–54, 2020. in the banana tree. The list of abbreviations is listed in M. G. Selvaraj, A. Vergara, H. Ruiz et al., “AI-powered banana diseases and pest detection,” Plant Methods, vol. 15, no. 1, Table 2. p. 92, 2019. The future work is to develop a system and implement J. Amara, B. Bouaziz, and A. Algergawy, “A deep learning- the proposed system in such a way that the disease can be based approach for banana leaf diseases classification,” in detected in banana cultivation in the agricultural farm with Datenbanksysteme für Business, Technologie und Web (BTW the aerial view drone camera instead of investigating an 2017) - Workshopband, B. Mitschang, D. Nicklas, F. Leymann individual plant. So that the intensity of the disease et al., Eds., pp. 79–88, Gesellschaft für Informatik e.V.., Bonn, spreading throughout the field can be easily estimated, 2017. which also helps in identifying the overall yield and loss in W. Liao, D. Ochoa, L. Gao, B. Zhang, and W. Philips, case of any disease infection, also furthermore a user- “Morphological Analysis for banana disease detection in close friendly mobile application will be created for the farmers range hyperspectral remote sensing images,” in Proceedings of to instantly identify the disease and react accordingly. the IGARSS 2019 - 2019 IEEE International Geoscience and [17–25]. Remote Sensing Symposium, pp. 3697–3700, Yokohama, Ja- pan, 28 July-2 August 2019. S. P. Deenan and J. SatheeshKumar, “Study on banana leaf Data Availability disease identification using image processing methods,” In- ternational Journal of Research in Computer Science and In- Data will be made available upon request formation Technology, vol. 2, pp. 89–94, 2014. B. Tigadi and B. Sharma, “Banana plant disease detection and Conflicts of Interest grading using image processing,” International Journal of Engineering Science and Computing, vol. 6, no. 6, 2016. The authors declare that they have no conflicts of interest. S. kaur, G. Babbar, and Gagandeep, “Image processing and classification, A method for plant disease detecion,” Inter- References national Journal of Innovative Technology and Exploring Engineering, vol. 8, no. 9S, 2019. F. Uwamahoro, A. Berlin, H. Bylund, C. Bucagu, and J. Yuen, J. Hang, D. Zhang, P. Chen, J. Zhang, and B. Wang, “Clas- “Management strategies for banana Xanthomonas wilt in sification of plant leaf diseases based on improved convolu- Rwanda include mixing indigenous and improved cultivars,” tional neural network,” Sensors, vol. 19, no. 19, p. 4161, 2019. Computational Intelligence and Neuroscience 13 G. Jayanthi, K. S. Archana, and A. Saritha, “Analysis of au- tomatic rice disease classification using image processing techniques,” International Journal of Engineering and Ad- vanced Technology, vol. 8, no. 3S, 2019. W.-J. Liang, H. Zhang, G.-F. Zhang, and H.-X. Cao, “Rice blast disease recognition using a deep convolutional neural network,” Scientific Reports, vol. 9, no. 1, p. 2869, 2019. I. Ahmad, M. Hamid, S. Yousaf, S. T. Shah, and M. O. Ahmad, “Optimizing pretrained convolutional neural networks for tomato leaf disease detection,” Complexity, vol. 2020, Article ID 8812019, 6 pages, 2020. A. Aruraj, A. Alex, M. S. P. Subathra, N. J. Sairamya, S. T. George, and S. E. V. Ewards, “Detection and classification of diseases of banana plant using local binary pattern and Support vector machine,” in Proceedings of the 2019 2nd International Conference on Signal Processing and Commu- nication (ICSPC), pp. 231–235, Coimbatore, India, March 2019. R. D. Devi, S. A. Nandhini, R. Hemalatha, and S. Radha, “IoT enabled efficient detection and classification of plant diseases for agricultural applications,” in Proceedings of the 2019 In- ternational Conference on Wireless Communications Signal Processing and Networking (WiSPNET), pp. 447–451, Chen- nai, India, March 2019. F. O. Sweetwilliams, V. O. Matthews, E. Adetiba, D. T. Babalola, and V. Akande, “Detection of Sigatoka disease in plantain using IoT and machine learning techniques,” Journal of Physics: Conference Series, vol. 1378, no. 2, Article ID 022004, 2019. W. Liao, D. Ochoa, Y. Zhao, G. M. V. Rugel, and W. Philips, “Banana disease detection by fusion of close range hyper- spectral image and high-resolution RGB image,” in Pro- ceedings of the IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, IEEE Valencia, Spain, July 2018. Q. Wang, F. Qi, M. Sun, J. Qu, and J. Xue, “Identification of tomato disease types and detection of infected areas based on deep convolutional neural networks and Object detection techniques,” Computational Intelligence and Neuroscience, vol. 2019, pp. 1–5, 2019. S. Kaur, S. Pandey, and S. Goel, “Plants disease identification and classification through leaf images: a survey,” Archives of Computational Methods in Engineering, vol. 26, no. 2, pp. 507–530, 2019. L. Li, S. Zhang, and B. Wang, “Plant disease detection and classification by deep learning-A review,” IEEE Access, vol. 9, pp. 56683–56698, 2021. Y. Guo, J. Zhang, C. Yin et al., “Plant disease identification based on deep learning algorithm in smart farming,” Discrete Dynamics in Nature and Society, vol. 2020, pp. 1–11, Article ID 2479172, 2020. A. Jan and G. M. Khan, “Real world anomalous scene de- tection and classification using multilayer deep neural net- works,” International Journal of Interactive Multimedia And Artificial Intelligence, pp. 1–10, InPress.