International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 11 (2019) pp.

2662-2666
© Research India Publications. http://www.ripublication.com

Farmer Buddy-Web Based Cotton Leaf Disease Detection Using CNN

Shantanu Kumbhar Amita Nilawar

Department Of Computer Engineering, Department Of Computer Engineering,
Pimpri Chinchwad College of Engineering, Pune, India. Pimpri Chinchwad College of Engineering, Pune, India.

Shruti Patil Bodireddy Mahalakshmi

Department Of Computer Engineering, Department Of Computer Engineering,
Pimpri Chinchwad College of Engineering, Pune, India. Pimpri Chinchwad College of Engineering, Pune, India.

Manasi Nipane
Department Of Computer Engineering,
Pimpri Chinchwad College of Engineering, Pune, India.

Abstract Neural Network and CNN. Previously image classification

algorithms like face recognition need to pay attention at where
Agriculture is one of the important professions in many
the face is located in an image this major problem is overcome
countries including India. As most part of the Indian financial
by CNN as well as features of an image are deeply processed
system is dependent on agriculture production, the keen
at each layer. Every disease on crop has different features
attention to the concern of food production is necessary. The
which are extracted at each layer of the convolution network.
taxonomy and identification of crop infection got much
importance in technical as well as economic in the The goal of this application is to develop a system which
Agricultural Industry. While keeping track of diseases in recognizes crop diseases. In this the user has to upload an
plants with the help of specialists can be very costly in image on the system, Image processing starts with the
agriculture region. There is a need for a system which can digitized color image of the diseased leaf. Finally by applying
automatically detect the diseases as it can bring revolution in the CNN plant disease can be predicted.
monitoring large fields of crop and then plant leaves can be
A. Purpose of Proposed System:
taken cure as soon as possible after detection of disease. The
1. Developing a user-friendly web-based system for
aim of the proposed system is to develop an application which
farmers
recognizes cotton leaf diseases. For availing this user need to
2. Recognizing Cotton leaf diseases accurately from
upload the image and then with the help of image processing
input images
we can get a digitized color image of a diseased leaf and then
3. Providing corrective and preventive measures for
we can proceed with applying CNN to predict cotton leaf
the detected diseases
disease.
Keywords: GLCM, CNN B. Cotton leaf diseases focused:
1. Alternaria Macrospora
2. Bacterial Blight
I. INTRODUCTION
We all know that most of the Indian land is used for II. RELATED WORK
agriculture purpose. It is also the back support for Indian
Siddharth Singh Chouhan et al.[1] gives Bacterial foraging
financial condition. There are various diseases which obstruct
optimization based Radial Basis Function Neural Network
the growth of crops in fields which may cause huge loss in the
(BRBFNN) for identification and classification of plant leaf
quality of products. Nowadays image processing is used a lot
diseases. Bacterial foraging optimization (BFO) method
for detecting such diseases, Pests like the germ, fungus, and
attains higher performance in terms of identification and
microorganisms are main cause of the disease to crops due to
classification of plant leaf diseases. The region growing
failure in excellence and extent of production. It is huge loss
algorithm increases the efficiency of the network by searching
to the farmer.
and grouping of seed points having common attributes for the
This paper is based on a system which implements feature extraction process.
Convolutional Neural Network to detect cotton leaf diseases.
Muhammad Waseem Tahir et al.[2] gives a novel fungus
It offers more capable ways to discover infection created by
dataset for the detection of fungus using CNN and distinguish
Bacteria and environmental effects. Disease detection at an
different types of fungus. The designed CNN architecture
early stage on crop is a challenging task for farmers where
gives 94.8% accuracy with 5 fold validation.
physical presence is a must. Disease detection and recognition
on the crop are very important. There are various algorithms Sukhvir Kaur et al. [3] provide a technique for the detection of
in image processing for disease recognition by image disease from Soybean leaf images. A rule-based semi-
classification like KNN, SVM, Random Forest, Artificial automatic system using concepts of k-means is designed and

2662
 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 11 (2019) pp. 2662-2666
© Research India Publications. http://www.ripublication.com

implemented to distinguish healthy leaves from diseased After prediction of plant disease system can also recommend
leaves. The experiment is performed on color features, texture pesticides to the farmers using android app. Android App is
features, and their combinations to train three models based also used to display the soil parameters values such as
on support vector machine classifier. Using SVM classify the humidity, moisture, and temperature along with the water
proposed system classify the leaf disease in 3 classes like i.e. level in a tank. By using Android app farmers can ON/OFF
downy mildew, frog eye, and sectorial leaf blight etc. The the relay to control the motor and sprinkler assembly
proposed system gives maximum average classification according to need.
accuracy reported is ~90% using a big dataset of 4775 images.
Ranjith et al.[4] provide smart irrigation system has been
III. METHODOLOGY
proposed which can control the irrigation automatically using
an android mobile application. Apart from this, the photos of The proposed system classifies the leaf image using image
plant leaves are captured and are sent to the cloud server, classification algorithm CNN. It can automatically detect and
which is further processed and compared with the diseased recognize diseases based on extracted features at each
plant leaf images in the cloud database. Based on the convolution layer. Below figure shows the structural design of
comparison a list of plant diseases suspected is given to the the planned system. The system used image processing
user via the Android mobile application. technique for disease detection. The user needs to upload the
cotton plant leaf image. The system can preprocess the
Adhao Asmita Sarangdharet al.[5] give Support Vector
uploaded image and then apply CNN technique. By using
Machine based regression system for identification and
CNN technique system can test the image with trained dataset
classification of five cotton leaf diseases i.e. Bacterial Blight,
and extract the features.
Alternaria, Gray Mildew, Cercospora, and Fusarium wilt.

Fig.1. System Architecture

A. Project Description: 1. Image Acquisition:

The techniques of machine vision are extensively applied to In this phase, raw image is taken as input from the user and
agricultural science, and it has great perspective especially in converted into equivalent gray scale image. Also the image is
the plant protection field, which ultimately leads to crops resized into size of 128*128.
management. The proposed system for plant leaf disease
2. Convolutional Layers:
detection is based on the infected images of various plants.
Images of the infected plants are captured by digital camera After the alteration of captured image, the processed image
and processed using image growing, image segmentation further passes through three different hidden layer in which
techniques to detect infected parts of the plants. feature extraction, pooling and flattening layer are also
performed.

2663
 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 11 (2019) pp. 2662-2666
© Research India Publications. http://www.ripublication.com

3. Disease Prediction: C. Implementation Details

After applying CNN, using Softmax layer the leaf image is Eclipse 3.3 Indigo IDE is used for overall development as a
predicted with disease which is gaining highest probability of standard platform. JSP is used to design GUI of the Web
occurrence. Application. The database is basically used for storing user
details like usernames, disease details, pesticide information
and pesticide costs. We used the MYSQL database as the data
B. Algorithm: to be stored was mostly structured. The tool used for DB
functionalities was MYSQL GUI Browser.
Traditional feature learning methods rely on semantic labels
of images as supervision. They usually assume that the tags By comparing different classification algorithm we come to
are evenly exclusive and thus do not point out towards the know that CNN is a feed-forward neural network that is used
complication of labels. The learned features endow explicit to analyze visual images by processing data. CNN can give a
semantic relations with words. higher accuracy in the prediction on plant disease due to
different advantages. We detect whether the plant is having
We also develop a novel cross-model feature that can both some disease or not and what pesticides are needed for
represent visual and textual contents. CNN itself is a recovery from that disease with the help of CNN. The datasets
technique of classifying images as a part of deep learning. In
are divided into Training and testing datasets in the ratio
which we apply single neural network to the full image. 70:30 and cross-validation is applied to achieve better
i. Accepts a volume of sizeW1×H1×D1 accuracy.
ii. Requires four hyper parameters:
For building CNN Python 3.6 is used along with libraries such
 Number of filters K as OpenCV for image processing and TensorFlow for
 Their spatial extent F Machine learning. Using libraries allows us to extend the
 The stride S functionalities for using it on larger database. CNN consists of
 The amount of zero padding P three Hidden layers i.e. Convolution layers; each convolution
iii. Produces a volume of size W2×H2×D2 where: layer consists of Pooling and Relu layers. Relu is the
a. W2=(W1-F+2P)/S+1 activation function used to remove the negative weights.
b. H2=(H1−F+2P)/S+1(i.e. width and height are compu MaxPooling is used in the pooling layers. 64 different filters
ted equally by symmetry) are applied in each layer for extracting features. Convolving
c. D2=K window size is 3*3 and stride size is 4.Softmax layer at the
iv. With parameter sharing, it introduces F*F*D1 weights end linearize the probabilities of the diseases in a single
per filter, for a total of (F*F*D1)*K weights and K dimensional matrix to choose maximum probability.
biases. In the output volume, the dth depth slice (of size
W2*H2) is the result of performing a valid convolution of Xampp server is used to deploy the Web Application on local
the dth filter over the input volume with a stride of S, and computer. Server handles the request and response to system
then offset by dth bias. user.
v. A common setting of the hyper parameters is F=3,
S=1,P=1
IV. RESULTS AND DISCUSSION
However, there are common conventions and rules of thumb
that motivate these hyper parameters. The goal of this application is to develop a system which
recognizes crop diseases and displays user the results as
detected disease, pesticides recommended and cost of
pesticides recommended, and for that user have to upload an
image then, Image processing starts with the digitized color
image of the diseased leaf. Finally by applying the CNN plant
disease can be predicted.
The dataset contains two types of diseased leaf images that are
Alternaria Macrospora and Bacterial Blight and also Healthy
Leaf images. The training dataset consists of total 513 images
while the testing dataset consists of 207 images. The accuracy
of training is 80% whereas the accuracy of testing is 89%.

Fig.2. Algorithmic Flow

2664
 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 11 (2019) pp. 2662-2666
© Research India Publications. http://www.ripublication.com

Fig.3. Output after prediction

Above figure show the actual output that is obtained after REFERENCES
classification of the disease. Also the disease description,
[1] S. Chouhan, A. Kaul, U. Singh, S. Jain,” Bacterial
cause of disease, pesticides needed, pesticides cost and the
foraging optimization based Radial Basis Function Neural
preventive measures are describes.
Network (BRBFNN) for identification and classification
of plant leaf diseases: An automatic approach towards
Plant Pathology”, 2017.
V. CONCLUSION AND FUTURE SCOPE
[2] M. W. Tahir, N. A. Zaidi, A. A. Rao, R. Blank, M. J.
A web-based system has been successfully implemented for
Vellekoop and W. Lang"A Fungus Spores Dataset and a
crop disease detection for cotton leaves using Convolutional
Convolutional Neural Networks based Approach for
Neural Network. Two diseases viz. Alternaria Macrospora and
Fungus Detection",IEEE TRANSACTION ON
Bacterial Blight are successfully being detected by system
NANOBIOSCIENCE, MAY 2018.
with training accuracy 80% using training dataset of 141
images for each disease. Healthy leaf image set of 141 images [3] S. Kaur , S. Pandey, S. Goel"Semi-automatic leaf disease
is also used for detecting healthy images. Convolutional detection and classification system for soybean
neural network has been developed with three hidden layers to culture",The Institution of Engineering and Technology
classify the cotton leaf disease images. System successfully 2018.
takes image input from user and provides input in the form of
[4] Ranjith, S. Anas, I. Badhusha, Zaheema OT, Faseela K,
disease detected, preventive measures, corrective measures,
M. Shelly"Cloud Based Automated Irrigation And Plant
pesticides required and probable cost for suggested pesticides.
Leaf Disease Detection System Using An Android
System can be extended to any other crops having availability Application",International Conference on Electronics,
of enough large dataset for that crop. Number of other diseases Communication and Aerospace Technology ICECA
can be included for detection. System also can implement 2017.
hardware using IOT for Image capturing in fields. The Web
[5] A. A. Sarangdhar, Prof. Dr. V. R. Pawar "Machine
interface can also involve a forum for formers to have
Learning Regression Technique for Cotton Leaf Disease
discussions regarding the current trends they are facing in
Detection and Controlling using IoT",International
different diseases.
Conference on Electronics, Communication and
Aerospace Technology ICECA 2017.

2665
 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 11 (2019) pp. 2662-2666
© Research India Publications. http://www.ripublication.com

[6] H. Waghmare, R. Kokare,” Detection and Classification

of Diseases of Grape Plant Using Opposite Colour Local
Binary Pattern Feature and Machine Learning for
Automated Decision Support System”, 2016 3rd
International Conference on Signal Processing and
Integrated Networks (SPIN).
[7] R. Sarkar, A. Pramanik,” Segmentation of Plant Disease
Spots Using Automatic SRG Algorithm: A Look Up
Table Approach”, 2015 International Conference on
Advances in Computer Engineering and Applications
(ICACEA) IMS Engineering College, Ghaziabad, India.
[8] Prof. U. Gawande , Mr. K. O. Hajari, Ms. K. R. Gavhale,
“Unhealthy Region of Citrus Leaf Detection Using Image
Processing Techniques”, International Conference for
Convergence of Technology – 2014.
[9] P. M. NesaRanil, T. Rajesh and R. Saravanan,”
Development of Expert System to Diagnose Rice
Diseases in Meghalaya State”, 2013 Fifth International
Conference on Advanced Computing (ICoAC).

2666