Plant Leaf Disease Detection using Deep Learning and CNN

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 394
Plant Leaf Disease Detection using Deep Learning and CNN
Sanjana Patel1
1Bengaluru,Karnataka,India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – All around the globe, problems such as forest
extinction and food shortage are major concerns for
economic development of a country and social welfare of the
people. There are different reasons for these issues. One of
the main reasons for these problems is diseases occurring in
the plant leaf. The potential threat to humans and other
animals because of the plant disease is immense. This paper
attempts to provide an end-to-end solution to automate the
detection of plant leaf disease and provides alert starting
from the source.
Through this paper we propose to create a novel low-cost
implementation of disease detection using state-of-the art
Deep learning concepts like Convolutional Neural Network,
Activation Functions, binary_crossentropy combined with
fundamental principles of RDBMS and UI/UX Our solution
aims to be a robust, low-cost, hardware-independent and
seamless solution for disease detection in present times. The
experimental accuracy of this system is 96.4% which is good
enough for real-world scenarios.
Key Words: Deep Learning, Convolutional Neural
Network, Binary Cross Entropy, RDBMS, Activation
Function
1. INTRODUCTION
Plant disease have impacted society and world history. All
varieties of plants wild and cultivated alike, are subject to
disease. The result of the plant losses due to disease can
result in hunger and starvation. Many valuable crops
might be very susceptible to disease and it might be a
possibility that survival of such plants without human
intervention is not possible. In such cases early detection
would be utmost important in order to prevent the further
spread of disease to other parts of the plant. The proposed
solution using deep learning techniques would be a vital
component to solve these issues as it does not involve any
hardware component and it is a seamless solution with
high accuracy to detect diseases in leaf as early as possible
and take appropriate measures.
It has become a very tedious problem for farmers to detect
the plant disease and treat it with appropriate measures
as early as possible. The current system that are in use are
ineffective and time-consuming. So using CNN based
model for disease detection would be a low-cost and an
effective approach to solve this problem.
In deep learning, a convolutional neural network (CNN, or
ConvNet) is a class of deep neural network, most applied
to analyse visual imagery. CNNs are more efficient than
other image classification algorithms. As they utilize little
pre-processing The learning is completely automated in
CNN
whereas in traditional algorithms these filters are hand-
engineered. This independence from prior knowledge and
completely automated without any human intervention in
feature extraction is a major advantage.
2. Literature Survey
Vijai Singh et all [1] describe the plants plays an important
role in agriculture field, as having disease in plants are
quite natural. This paper subtly informs that proper care
should be taken in certain area else it causes serious
effects in plants. This paper proposes an algorithm for
image segmentation which is used for automatic detection
and classification of plant leaf diseases.
Amar Kumar Deya et all [2] discuss leaf rot disease
detection for betel vine (Piper betel L.) based on image
processing algorithm. This paper presents the drawback of
current manual detection and uses identifying color
feature of the rotted leaf areas to find detection in plant
diseases. Subsequently, the rotted area was segmented,
and area of rotted leaf portion was extracted from the
various plant feature data. The results showed a
performance of this automatic vision-based system in
practice with easy validation.
A.P Soni et all [3] Proposed a green heart shaped betel leaf,
in India it is known as Paan. This paper covers the easy,
accurate, and less human intervention method of leaf area
measurement. Leaf area of plants is a useful tool in
physiological and agronomic studies. Around 1000 betel
leaf investigation is performed and the paper includes
important aspects of these investigation. These results are
in turn compared with the graphical technique of leaf area
measurement. The advantage of this method is the
easiness and the accuracy of finding the estimated leaf
area precisely.
Daisy shergill& et.al.,[4] describes a approach is useful in
crop protection especially in large area farms, which is
based on automated techniques that can detect diseased
leaves using color information of leaves. The disease can
be detected by capturing an image of a certain plant leaf

followed by extracting feature from the captured image.
First the captured RGB image is converted to gray image
and then gray image is resized and perform canny edge
detection, apply various comparison techniques, which
detect the presence of disease and also the type of
diseases. It enables to find the disease early so that
appropriate measures can be taken.
RenukaRajendraKajale [5] describes the approach for
detection and computation of texture information for
plant leaf diseases. The processing system consists of four
main steps, color image is converted to HSI, then the green
pixels are masked and removed using specific threshold
value, then the preprocessed image is divided, and the
useful segments are extracted, finally most important
information which is texture is obtained in this algorithm.
obtained. The diseases present on the plant leaf are
evaluated based on the texture information.
Prakash M. Mainkar& et.al.,[6] provides a software
solution to automatically detect and classify plant leaf
disease. This approach will increase productivity of crops.
It includes series of steps which are acquisition of image,
image preprocessing, segmentation, feature extraction and
classification.
Niket Amoda& et.al., [7] provide image processing-based
solutions that are highly automatic, less cost and accurate.
Solution is composed of four main steps; in the first step
the RGB leaf image is transformed to other color model.
Next, in the second step, the images which are
transformed are segmented to obtain the most important
information. In the third step, the calculation is done
based on the features of texture of leaf.. Finally, in the
fourth step the classification is done by using pre-trained
neural network based on the result of feature extraction.
Overall, the approach of training deep learning models on
increasingly large and publicly available image datasets
presents a clear path towards smartphone-assisted crop
disease diagnosis on a massive global scale.
3. Proposed System
The Proposed system opted to develop is an algorithm
used to recognize diseases in crop leaves by using
convolutional neural network. Devices like digital camera
or any identical device are used to take the pictures of
affected part of plant, these images are then used to
identify the affected area in leaves. Finally image-
processing techniques are applied on them, to process
those images and obtain unique and important features
needed to analyze later. This model is properly trained
using CNN and then the classification takes place. Then the
comparison of the test image and trained model takes
place followed by the display of the result. If there is a
defect or disease in the plant the software displays the
result
4. System Architecture
A convolutional neural network is made up of input layer,
hidden layers and an output layer. In a feed-forward
neural network, these middle layers are being addressed
as hidden layers because the inputs being fed to hidden
layers and output generated from them are masked by
activation function and final convolution. In a
convolutional neural network, the hidden layers are most
essential as they perform convolutions. This Conv layer
does most of the heavy lifting. The output from Conv layer
is termed as Feature Map which gives us the important
information of the image. Following these, CNN consists of
layers like Pooling layers, fully connected layers and
normalization layers which will help to process the image
and provide results more accurately. The primary
application of pooling layer is to reduce the computational
cost by decreasing the size of convolved feature map. The
fully connected layer comes before the output layer which
consists of weights and biases and all the main
mathematical computation takes place in this layer. In this
layer, the main function that is classification takes place.
These layers along with additional features like dropout
and Activation function makes the model more efficient
5. System Implementation

5.1 Preprocessing
The input image captured is preprocessed by following
steps. After reading the image the next step is to resize the
image. The resizing of the image is vital because the
images obtained from the dataset will be of different sizes
which in turn would be difficult to process through the
model. Therefore, we must make sure that a base size is
set for all images and then fed into our AI algorithms. Then
we try to remove the blur by reducing the noise. In the
next step, we step we are going to segment the image,
separating the background from foreground objects and
we are going to further improve our segmentation with
more noise removal. The preprocessing step helps to
reduce the unwanted section and focuses on the main
object in the image.
5.2 Feature Extraction
Features are patterns of an object in an image which will
help to identify the desired result. Features include
different types of properties such as corners, edges,
regions of interest points, ridges, etc. This method will be
most useful when the dataset is large and because of this
technique it is possible to reduce the number of features
without losing any important or relevant information. So,
by this processing gets easier. Convolutional Neural
Network will be able to extract complex features
compared to Traditional feature extractors where CNN are
able to express the image in much more detail and are
very efficient. These large datasets have multiple different
variables. These variables require high computing
resources in order to process them. So, Feature extraction
identifies the best feature from those big data sets thus,
effectively reducing the amount of data. Ideally, these
features are relatively easy to process, but still are able to
describe the actual data set with the accuracy and
precision. The Convolution layers and pooling layers helps
us to extract features from an image. The amount of
redundant data will be significantly eliminated with the
help of feature extraction. In the end, the reduction of the
data helps to build the model with low-cost and more
efficiently and also increase the speed of learning and
generalization steps in the machine learning process.
5.3 Training the Neural Network
The training set consist of n number of images stored in
the database where both plants infected with disease and
plants not infected with disease are present. Since this is a
binary classification, we can use binary_crossentropy. The
main aim of the training model is to reduce the loss and in
order to achieve this we pass the optimizer parameter.
There are many options that can be used. To obtain high
accuracy, metric model is deployed where the parameters
like number of epochs, model loaders etc. are used to
make the model more precise and accurate. We pass the
image(data) from the training dataset (to the model. Then
calculate the loss and do a backpropagation during which
we update the weights based on the loss. Then the image
is fed, and the model detects whether the plant is infected
with a disease or not.
5.5 Result
● Using this method, the plant disease can be
identified at an early stage. The images will be
processed quickly, and results can be obtained
within a short span of time.
● This method mainly focuses on large databases
and advance feature of color extraction that
provides the better result of detection.
● The suitable prevention measures are provided to
the farmers based on detection of disease.
● Training samples have been increased to improve
the efficiency of disease detection.
6 Conclusion
The proposed system was developed for the welfare of the
farmers and provide a boost in agricultural sector. The
developed system can detect disease in plant early and
help to take measures to avoid the spread of disease to
other parts of the plant. By proper knowledge of the
disease the remedy can be taken for improving the health
of the plant This paper presents a basic overview of the
concept of CNN and underlines the application of CNN in
plant leaf disease detection, high accuracy, and low cost,
thereby providing a more control and ease of disease
detection. Moreover, it highlights the problems in older
technology and puts some light on recent developments of
CNN architecture in deep learning.
REFERENCES
[1] Diseased plant leaves using Neural Network
Algorithms K. Muthukannan1, P. Latha2, R. Pon Selvi1 and
P. Nisha1.
[2] Detection and Classification of Leaf Disease using
Artificial Neural Network Malvika Ranjan1, Manasi Rajiv
Weginwar2, Neha Joshi3, Prof. A.B. Ingole4 International
Journal of Technical Research and Applications.
[3] Deep Neural Networks Based Recognition of Plant
Diseases by Leaf Image Classification Dubravko Culibrk,2
and Darko Stefanovic1
[4] Plant Disease Classification Using Convolutional
Networks and Generative Adverserial Networks Emanuel
Cortes Stanford University.
[5] A Robust Deep-Learning-Based Detector for Real-Time
Tomato Plant Diseases and Pests Recognition , Alvaro

Fuentes 1, Sook Yoon 2,3, Sang Cheol Kim 4 and Dong Sun
Park Sensors 2017, 17, 2022; doi:10.3390/s17092022.
[6] Identification of Apple Leaf Diseases Based on Deep
Convolutional Neural Networks, Bin Liu 1,2,*,† Yun Zhang
1,†, DongJian He 2,3 and Yuxiang Li 4, Symmetry 2018, 10,
11; doi:10.3390/sym10010011.
[7] Using Deep Learning for Image-Based Plant Disease
Detection Sharada Prasanna Mohanty1,2, David
Hughes3,4,5, and Marcel Salathé1,2,6 April 15, 2016 | 1
https://arxiv.org/pdf/1604.03169.
[8] Image-Based Plant Disease Detection with Deep
Learning Ashwin Dhakal1, Prof. Dr. Subarna Shakya2 ISSN:
2231 – 2803 http://www.ijcttjournal.org

Plant Leaf Disease Detection using Deep Learning and CNN

More Related Content

Similar to Plant Leaf Disease Detection using Deep Learning and CNN (20)

More from IRJET Journal (20)

Recently uploaded (20)

Plant Leaf Disease Detection using Deep Learning and CNN