9 VI June 2021

https://doi.org/10.22214/ijraset.2021.35040
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue VI Jun 2021- Available at www.ijraset.com

Image Processing based Plant Disease Detection

using CNN
Prof. Rachana Sable1, Omkar Gawali2, Mangesh Bhise3, Vaibhav Dond4, Saurabh Jadhav5
1, 2, 3, 4., 5
Department of Computer Engineering, G.H. Raisoni Institute of Engineering and Technology, Wagholi, Pune, Maharashtra,
India.

Abstract: In the era of Scientific Development, many technologies and new ways of solving real-life problems are being invented
every day. The basic need of the food is increasing parallelly, due to an increase in population. According to FAO of the UN,
annually 20 to 40 percent of crops were lost due to diseases. That’s why technological development in the agricultural field is
important to improve the productivity of crops. This major issue can be overcome by implementing disease detection techniques
to identify the disease from an input image. This process involves steps like dataset collection, image pre-processing and training
a classification model. The dataset consists of plants like cotton, grapes and tomatoes. CNN classification model is used for
disease classification. The proposed model gives an accuracy of 96%. After disease classification it also provides information like
causes, symptoms, management and diagnostic solutions.
Keywords: Identification of diseases, Deep learning, Solutions, Convolution Neural Network (CNN)

I. INTRODUCTION
Agriculture is a vital component of the global economy, particularly in underdeveloped countries. It is the primary source of
employment, income, and food, and these basic needs are fulfilled by agriculture all over the world. According to the Food and
Agriculture Organization (FAO), the share of the agricultural population is 67% of the total population. 58% of indian population is
dependent on agricultural practices.
According to the Food and Agriculture Organization of the United Nations (FAO), up to 40% of food crops worldwide are lost
every year due to pests and plant diseases. The farmers use traditional naked eye observation techniques to identify diseases in
plants. Because of this traditional technique most of the time famers cannot identify the correct disease which may cause loss in
productivity of crops. If no immediate action is taken by farmers it is very difficult to manage the diseases in crops. Farmers are also
unaware of many diseases of plants. Hiring an expert for this purpose is not affordable for them. So in this paper a model has been
proposed to identify the diseases in plants from an input image.
The proposed system will detect the diseases of plants like cotton,grapes and tomato through image processing where farmers will
have to provide image of the disease affected leaves, where initially the image will undergo various steps like resizing, rescaling,
rotating(if required), etc.After that the processed image will be passed to the classification model, where the trained model will
identify the disease.The two main things which needs to be focused are accuracy and speed.
Here the proposed system uses the CNN model for classification. In this work a total of 7,800 leaf images are used for model
training. It consists of 4 diseases of Cotton plants, 3 diseases of grapes, 9 diseases of Tomato plants and healthy leaves of each plant.

II. LITERATURE REVIEW

Paper by A. Jenifa, R. Ramalakshmi and V. Ramachandran gives solutions to the plant disease with image classification. In their
approach they collect 500 images for training and 100 images for testing the model. There were a total of 5 classes that include 4
disease classes and one normal healthy leaf class. Removal of noise is done with some image preprocessing and then conversion
into lab color model was done.Classification is done using Deep Convolutional Neural Network. They have achieved 96% as
average accuracy [1].
Another paper named “Classification Of Plant Leaf Diseases Using Machine Learning And Image Preprocessing Techniques”
clarifies that they have used Gaussian Blur for image-preprocessing, k-means segmentation, CNN model for the leaf disease
classification.In their methodology they have used a dataset of 20,000 images divided into 19 classes.They achieved accuracy 66.4%
using logistic regression, 54.5% using KNN , 53.4% using SVM and 98% using CNN. Their proposed model was for tomato and
grapes related diseases [2].

©IJRASET: All Rights are Reserved 656

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue VI Jun 2021- Available at www.ijraset.com

Another paper by S. V. Militante, B. D. Gerardo and N. V. Dionisio used the CNN for the classification purpose.Image dataset was
gathered from the plant village repository. It consists of 35,000 images with 32 classes of diseases.Their proposed model was for
tomato and grapes related diseases.They have achieved average accuracy of 96.5% [3].
Paper by N. R. Bhimte and V. R. Thool has taken a cotton plant for disease detection. In their approach, the collection contains 130
photos, including 50 photos of Bacterial blight, 50 images of Magnesium shortage, and 30 images of healthy leaves, according to
their method. They have used k-means for segmentation, GLCM for feature extraction and CNN model for the leaf disease
classification. They have achieved 98.46% as average accuracy [4].
R. G. de Luna, E. P. Dadios and A. A. Bandala’s “Automated Image Capturing System for Deep Learning-based Tomato Plant Leaf
Disease Detection and Recognition” uses faster R-CNN and CNN for the classification purpose.The dataset consists of 4,923 images
of tomato plant leaves which are categorized into Healthy and diseased leaves..Their proposed model was only for tomato related
diseases.They have achieved accuracy of 91.67% using CNN and 80% using F-RCNN[5].
“Plant Leaf Disease Detection and Classification using Image Processing” clarifies that they have used k-means segmentation and
SVM for the leaf disease classification.In their methodology dataset consist of total 560 images divided into 4 classes. They
achieved average accuracy 98%.Their proposed model was for grapes and cotton related diseases [6].
“Grape Leaf Disease Identification Using Improved Deep Convolutional Neural Networks” clarifies that they have used DICNN
and SVM for the leaf disease classification.In their methodology dataset consist of total 7,669 images divided into 5 classes.They
achieved accuracy of 67.82% using SVM and 97.22% using DICNN.Their proposed model was only for grapes related diseases [7].
N. Agrawal, J. Singhai and D. K. Agarwal proposed to have image processing techniques used in the early detection of plant
diseases through leaf features inspection. They mentioned implementing image analysis and classification techniques for the
extraction and classification of leaf diseases. Furthermore, the authors focused on image preprocessing, they transformed the image
into the LAB color model and also the HSI color model and the classification is done well using SVM. The given system using
features from both LAB and HSI color model gives 90% average accuracy for all three-diseased leaf and healthy leaf compared to
82.5% in case of taking only LAB color model for feature extraction [8]

III. PROPOSED SYSTEM

A. Architecture

Fig No. 3.1 - System Architecture

©IJRASET: All Rights are Reserved 657

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue VI Jun 2021- Available at www.ijraset.com

To classify and find the type of disease in plants, our proposed model consists of following stages:
1) Collection / Gathering of Dataset: We have gathered the images of cotton plants by visiting the field and capturing the images
of leaves. The dataset of grapes and tomato has been collected from kaggle. The dataset consists of 7,800 leaf images
containing 19 classes.
2) Image Preprocessing: Image Preprocessing are the steps taken to preprocess the images into standard format. It helps to
increase model execution speed and to reduce model training time. In this system ‘Keras' library is used for image
preprocessing. The image pre-processing consists of resizing , rescaling , removing noise, shearing, rotating, zooming and
flipping .Images are resized to 150 x 150 resolution for processing.
3) Image Classification: For disease classification Deep Convolutional Neural Network (CNN) is used. The CNN model consists
of 4 layers
a) Convolution Layer
b) Pooling Layer
c) Dense Layer
d) Output Layer

The activation function used in the proposed model are :

 ReLU
 Softmax for output layer

IV. SYSTEM IMPLEMENTATION

The website consists of 2 modules:

A. User Panel
This module consists of submodules which will provide various functionalities to user like :
1) Register him/her to the website. Login by using credentials.
2) If the user wants to detect the disease of the plant then he needs to follow the steps likes first select the type of plant whose
disease need to be detect.

Fig. No. 4.1 - Crops Selection

3) Then upload the image of the plant leaf.

4) Then the user has to click on the predict button If the system predicts that it is an infected leaf, then our system will show the
type of disease detected ,it’s causes , symptoms, precautions need to be taken and solutions to cure the disease. Otherwise the
system will show it is a healthy leaf
5) If the user wants some expert advice then he can post queries on the portal.
6) Users can also view the agricultural posts and weather forecasting details.

Fig. No. - 4.2 - Upload Image

©IJRASET: All Rights are Reserved 658

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue VI Jun 2021- Available at www.ijraset.com

Fig. No. 4.2 - Predicted Disease

B. Admin and Support Panel

This module consists of submodules which will provide various functionalities to admin like :
1) View / Add / Delete user profile.
2) Modify / Add new plant species.
3) Modify / Add diseases information of a plant
4) Display and provide support to user queries
5) Add / Update / Delete informative news and post for the users
6) Admin can also view the weather forecasting details.

V. EXPERIMENTAL ANALYSIS
A. Cotton Plant
Tested for 10 images of each disease.
Table 5.1 – Cotton Plant
Sr. No. Disease name Accuracy

1. Alternaria 90%

2. Bacterial blight 80%

3. Grey Mildew 90%

4. Chlorosis 90%

5. Healthy 100%

B. Grapes Plant
Tested for 10 images of each disease.
Table 5.2 – Grapes Plant
Sr. No. Disease name Accuracy

1. Black rot 90%

2. Leaf blight 100%

3. Black Measles 100%

4. Healthy 100%

©IJRASET: All Rights are Reserved 659

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue VI Jun 2021- Available at www.ijraset.com

C. Tomato Plant
Tested for 20 images of each disease.
Table 5.3 – Tomato Plant
Sr. No. Disease name Accuracy

1. Late blight 80%

2. Leaf Mold 65%

3. Septoria Leaf spot 70%

4. Target spot 80%

5. Bacterial spot 95%

6. Early blight 75%

7. Mosaic virus 85%

8. Spider mites 65%

9. Yellow Leaf Curl 95%

virus

10. Healthy 90%

VI. CONCLUSION
In this paper, a web application has been implemented for disease classification of multiple plants. The proposed CNN model has
been trained for diseases like Cotton, Grapes and Tomatoes. The input image has been preprocessed and then passed to the
classification model. The major diseases occurring in this plants are taken into consideration. The average accuracy of cotton plant
is 90%, grapes plant is 97.5% and tomato plant is 80%.After the disease prediction the web application also provides information
about the disease like causes , symptoms, preventive measures and diagnostic solution for disease.

VII. FUTURE SCOPE

In future work, we would like to add a new plant and its diseases to web applications to enhance its functionality. Provide a good
marketing environment for the farmers to help them to gain good agricultural income. Provide information and notifications about
government schemes .
REFERENCES
[1] Jenifa, R. Ramalakshmi and V. RamachandranR. Ramalakshmi, "Cotton Leaf Disease Classification using Deep Convolution Neural Network for Sustainable
Cotton Production," 2019 IEEE International Conference on Clean Energy and Energy Efficient Electronics Circuit for Sustainable Development (INCCES),
2019, pp. 1-3, doi: 10.1109/INCCES47820.2019.9167715.
[2] P. Sharma, P. Hans and S. C. Gupta, "Classification Of Plant Leaf Diseases Using Machine Learning And Image Preprocessing Techniques," 2020 10th
International Conference on Cloud Computing, Data Science & Engineering (Confluence), 2020, pp. 480-484, doi: 10.1109/Confluence47617.2020.9057889.
[3] S. V. Militante, B. D. Gerardo and N. V. Dionisio, "Plant Leaf Detection and Disease Recognition using Deep Learning," 2019 IEEE Eurasia Conference on
IOT, Communication and Engineering (ECICE), 2019, pp. 579-582, doi: 10.1109/ECICE47484.2019.8942686.
[4] N. R. Bhimte and V. R. Thool, "Diseases Detection of Cotton Leaf Spot Using Image Processing and SVM Classifier," 2018 Second International Conference
on Intelligent Computing and Control Systems (ICICCS), 2018, pp. 340-344, doi: 10.1109/ICCONS.2018.8662906.
[5] R. G. de Luna, E. P. Dadios and A. A. Bandala, "Automated Image Capturing System for Deep Learning-based Tomato Plant Leaf Disease Detection and
Recognition," TENCON 2018 - 2018 IEEE Region 10 Conference, 2018, pp. 1414-1419, doi: 10.1109/TENCON.2018.8650088.
[6] Oo, Yin & Htun, Nay. (2018). “Plant Leaf Disease Detection and Classification using Image Processing”. International Journal of Research and Engineering. 5.
516-523. 10.21276/ijre.2018.5.9.4.
[7] Liu Bin, Ding Zefeng, Tian Liangliang, He Dongjian, Li Shuqin, Wang Hongyan, TITLE=Grape Leaf Disease Identification Using Improved Deep
Convolutional Neural Networks, JOURNAL=Frontiers in Plant Science, VOLUME=11, YEAR=2020,
[8] N. Agrawal, J. Singhai and D. K. Agarwal, "Grape leaf disease detection and classification using multi-class support vector machine," 2017 International
Conference on Recent Innovations in Signal processing and Embedded Systems (RISE), 2017, pp. 238-244, doi: 10.1109/RISE.2017.8378160.

©IJRASET: All Rights are Reserved 660