Convolutional Neural Network for Leaf and citrus fruit disease identification using deep learning model

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 390
Convolutional Neural Network for Leaf and citrus fruit disease
identification using deep learning model
Ragavi .R1, Kaviya Sri .S2, Pooja Rao .P3, Ms. Thenmozhi .R AP (Sel.G)4
1,2,3 SRM Valliammai Engineering College, Kattankulathur, Chennai
4 Associate Professor, Department of Information Technology, SRM Valliammai Engineering College, Tamil Nadu,
India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Grouping of the different sorts of citrus organic
products in light of their skin tone is an arduous errand and
consumes a ton of time. An automated framework whenever
prepared to arrange citrus organic products in light of their
external skin tone, it would be of extraordinary assistance
for the ventures that perform arranging and evaluating of
the farming items.
The target giving a viewpoint to the arrangement of
seven various types of citrus leafy foods by a prepared
computerized framework. The framework is prepared to
such an extent that it would anticipate the biggest maker of
the concerned citrus organic product on the planet and will
likewise pass on a message containing the supplement worth
or medical advantages of the particular citrus organic
product.
The pre-prepared convolution brain network models
like AlexNet and LeNet Architecture when retrained with a
dataset of 2029 pictures gave us attainable outcomes. The
exploratory outcomes as exactness and misfortune rate
displayed in the charts approves the adequacy of the
proposed methods.
Keywords : Citrus Fruits & Leaves Classification, Neural
Network, Tensorflow
I. INTRODUCTION
Farming examination means to increment food
creation and quality while bringing down costs and
helping benefits. Natural product trees assume a
significant part in any state's monetary turn of events. One
of the most notable natural product plant species is the
citrus plant, which is high in L-ascorbic acid.
Citrus organic products are known as agrumes, which
generally signifies "harsh organic products". Most citrus
natural products are from a little gathering of plants that
are presumably gotten from around four species in
Southeast Asia. The most widely recognized are oranges,
grapefruits, tangerines, lemons, and limes. Citrus organic
products contain many mixtures that can assist with
keeping your heart sound. Their dissolvable fiber and
flavonoids may assist with raising solid HDL cholesterol
and lower unsafe LDL cholesterol and fatty substances.
The natural products might bring down hypertension,
another gamble factor for coronary illness.
Convolutional brain networks are profound learning
calculations that are exceptionally strong for the picture
handling is utilized to examination the leaf and natural
product qualities. The rural area is a huge and new system
for scientists in the field of PC vision today. Agribusiness'
essential objective is to create a wide scope of important
and significant harvests and plants. In cultivating, plant
microorganisms lessen the sum and attributes of the item,
so they should be controlled early.
As of late, rural analysts have zeroed in their
endeavors on infections of different foods grown from the
ground. The analysts concocted a few strategies for
distinguishing and grouping infections in foods grown
from the ground . Pre-handling and side effect division are
finished utilizing an assortment of new strategies.
II. LITERATURE SURVEY
[1] The proposed CNN model extracts complementary
discriminative features by integrating multiple layers. The
CNN model was checked against many state-of-the-art
deep learning models on the Citrus and Plant Village
datasets. According to the experimental results, the CNN
Model outperforms the competitors in a variety of
measurement metrics. The CNN Model has a test accuracy
of 94.55 percent, making it a valuable decision support
tool for farmers looking to classify citrus fruit/leaf
diseases.

[2] The use of deep-learning models (AlexNet, VggNet,
ResNet) pre-trained on object categories (ImageNet) in
applied texture classification problems such as plant
disease detection tasks. The proposed method proves to be
significantly more efficient in terms of processing times
and discriminative power, being able to surpass traditional
and end-to-end CNN-based methods and provide a
solution also to the problem of the reduced datasets
available for precision agriculture.
[3]A novel deep learning (DL) based citrus disease
detection and classification model. A new DL based
AlexNet architecture is employed for effective
identification of diseases. The presented model involves
four main processes namely pre-processing, segmentation,
feature extraction, and classification.
[4]Agriculture has a major role in the economic
development of our country. Productive growth and high
yield production of fruits is essential and required for the
agricultural industry. Application of image processing has
helped agriculture to improve yield estimation, disease
detection, fruit sorting, irrigation and maturity grading.
Image processing techniques can be used to reduce the
time consumption and has made it cost efficient. In this
paper, we have provided a survey to address these
challenges using image processing techniques.
[5] Machine learning based approach is presented for
classifying and identifying 10 different fruit with a dataset
that contains 6847 images use 4793 images for training,
1027 images for validation and 1027 images for testing. A
deep learning technique that extensively applied to image
recognition was used. We used 70% from image for
training and 15% from image for validation 15% for
testing. Our trained model achieved an accuracy of 100%
on a held-out test set, demonstrating the feasibility of this
approach.
[6] Deep convolutional-neural-network (CNN) models are
implemented to identify and diagnose diseases in plants
from their leaves, since CNNs have achieved impressive
results in the field of machine vision. Standard CNN models
require a large number of parameters and higher
computation cost. In this paper, we replaced standard
convolution with depth=separable convolution, which
reduces the parameter number and computation cost. To
evaluate the performance of the models, different
parameters such as batch size, dropout, and different
numbers of epochs were incorporated. In comparison with
other deep-learning models, the implemented model
achieved better performance in terms of accuracy and it
required less training time
[7] The production of citrus fruits. Lemon, Oranges and
Mosambi in India is around 13% in overall production of
fruits. Wastage in post harvesting is around 20% of its
production as they are not ripened. Non-climacteric fruits,
once4 harvested, never ripen further.
[8] It is mainly focused on to increase yield wherein
Agriculture the automatic methods for counting the
number of fruits play a critical role in crop management. In
this paper a review of previous studies and systems to
count the number of fruits on trees and their yield
estimation is performed. The various computer vision and
optimization techniques are presented to automate the
process of counting fruits. Further the main features and
drawbacks of the previous systems in this area are
summarized in this paper.
III. PROPOSED METHODOLOGIES
In the creation of organic products , they likewise get
impacted by sickness which brings about diminished
return and benefit. These days use of pesticides, composts
and bug sprays , has raised .Even however utilizing them
influences the tree involving them in perfect sum and
when required is vital to note. For utilizing composts that
have synthetic substances as well as for it are normally
made to utilize excrement that. To make it understood,
treating the patient without realizing the sickness will
likewise bring about infection.
Subsequently, recognizable proof of illness of trees
and plants are truly significant. So here mostly citrus
natural products are made into center, as they are plentiful
in Vitamin-C and furthermore has a few health advantages.
One such essential to make reference to is that it keeps
heart sound as they contain fiber and flavonoids might
assist with raising solid HDL cholesterol and lower unsafe
LDL cholesterol and fatty substances. And furthermore,
they lower hypertension.
In the proposed framework, utilizing Convolutional Neural
Network the infection of the Citrus natural product is
recognized. In this organization model, the dataset is
shipped off convolutional layers, pooling layers, and
completely associated (FC) layers. For this undertaking we
have utilized 2788 dataset from Kaggle to recognize 7
explicit infections of citrus organic product specifically,
Canker, Greening, Healthy, Limon Corillo, Limon
Mandarino, Mandarina Israeli, Mandarina Pieldesapo. To

group the citrus leafy foods. We wanted to configuration
profound learning method so an individual with lesser
aptitude in programming ought to likewise have the option
to handily utilize it. The proposed framework to
anticipating citrus foods grown from the ground. It makes
sense of about the exploratory investigation of our
procedure. Different number of pictures is gathered for
every classification that was arranged into dataset pictures
and info pictures. The essential credits of the picture are
depended upon the shape and surface arranged highlights.
The example screen captures show the citrus foods grown
from the ground utilizing TensorFlow model. Exactness we
got is 98%.
ADVANTAGES
In the proposed structure, utilizing Convolutional Neural
This for the most part centers around expanding
throughput and decreasing emotion emerging from human
specialists in recognizing the citrus natural products and
leaf. And furthermore carried out multiple Architectures
for getting more exactness.
In this we are conveying in a page utilizing python's
Django structure. Subsequently it would be more
proficient to utilize.
In this work, we have involved CNN calculation as
contrasted and past fake organization technique profound
learning is more exact in acknowledgment and picture
arrangement.
PROBLEM IDENTIFICATION
The objective is to recognize the illness of the citrus plant
with more accuracy.Earlier in the business related to this,
they have executed it utilizing a solitary calculation and
architecture.We utilized 2 engineering in particular
AlexNet and LeNet structure to obtain results more exact.
To recognize the sickness we really want more number of
dataset, yet we dint have enough dataset to prepare the
model.Hence we wound up looking and catching pictures
of them and uploading them.
IV. ARCHITECTURE DIAGRAM
In the above diagram, initially we collect data and store it.
Then we test and train the data, next it is sent to image
augmentation section where the images undergo process
like flip , flop, rotate etc. in order to get accurate results.
Once the image gets augmented, from that its feature gets
extracted and the required feature are selected and used
for evaluating.
Now comes the important part of the project, where the
CNN algorithm gets implemented. By using CNN algorithm,
the image gets passed through four layers [2]. From this
we can classify the disease and identify. Now the model is
completely ready to get deployed. Finally the input from
the users are obtained and then processed and evaluated.
From the processing and evaluating, output is obtained.
V. MODULES
MANUAL NET
Import the given image from dataset:
Need to import our informational index utilizing keras
preprocessing picture information generator work
likewise we make size, rescale, range, zoom range, level
flip. Then we import our picture dataset from organizer
through the information generator work. Here we set
train, test, and approval additionally we set target size,
group size and class-mode from this capacity we need to
prepare utilizing our own made organization by adding
layers of CNN.

Diseases are:
CANKER
GREENING
HEALTHY
LIMON CRIOLLO
LIMON MANDARINO
MANDARINA ISRAELI
MANDARINA PIELDESAPO
 To train the module by given image dataset
To prepare our dataset utilizing classifier and fit
generator work additionally we make preparing steps per
age's then absolute number of ages, approval information
and approval steps utilizing this information we can
prepare our dataset.
 Working process of layers in CNN model
A Convolutional Neural Network (ConvNet/CNN)
is a Deep Learning algorithm which can take in an input
image, assign importance (learnable weights and biases) to
various aspects/objects in the image and be able to
differentiate one from the other.

The pre-taking care of expected in a ConvNet is a great
deal of lower when diverged from other portrayal
estimations. While in rough methods channels are hand-
planned, with enough arrangement, ConvNets can get
comfortable with these channels/ascribes. The plan of a
ConvNet is basically identical to that of the accessibility
illustration of Neurons in the Human Brain and was
charged up by the relationship of the Visual Cortex.
Individual neurons answer supports simply in a restricted
region of the visual field known as the Receptive Field.
Their association involves four layers with 1,024
information units, 256 units in the essential mystery layer,
eight units in the subsequent mystery layer, and two
outcome units.
ALEXNET
AlexNet is the name of a convolutional brain network
which generally affects the field of AI, explicitly in the
utilization of profound figuring out how to machine vision.
AlexNet was the first convolutional network .
AlexNet engineering comprises of 5 convolutional
layers, 3 max-pooling layers, 2 standardization layers, 2
completely associated layers, and 1 SoftMax layer. Each
convolutional layer comprises of convolutional channels
and a nonlinear initiation work ReLU. The pooling layers
are utilized to perform max pooling.
Convolutional layers
Convolutional layers are the layers where filters are
applied to the original image, or to other feature maps in a
deep CNN. This is where most of the user-specified
parameters are in the network. The most important
parameters are the number of kernels and the size of the
kernels.
Pooling layers
Pooling layers are like convolutional layers, yet they fill a
particular role, for example, max pooling, which takes the
greatest worth in a specific channel district, or normal
pooling, which takes the typical worth in a channel area.
These are regularly used to lessen the dimensionality of
the organization.
Dense or Fully connected layers
Fully connected layers are put before the
characterization result of a CNN and are utilized to
straighten the outcomes before arrangement. This is like
the result layer of a MLP.
LENET
LeNet was one among the earliest convolutional brain
networks which advanced the occasion of profound
learning. After innumerous long periods of investigation
and a lot of convincing emphasess, the final product was
named LeNet..
Architecture of LeNet-5:
LeNet-5 CNN design is comprised of 7 layers. The layer
organization comprises of 3 convolutional layers, 2
subsampling layers and 2 completely associated layers.

DEPLOY
Conveying the model in Django Framework and
anticipating yield
In this module the prepared profound learning model is
changed over into progressive information design
document (.h5 record) which is then sent in our Django
system for giving better UI and foreseeing the result
whether the given RGB pictures is
CANKER/GREENING/HEALTHY/LIMON CRIOLLO/LIMON
MANDARINO/MANDARINA ISRAELI/MANDARINA
PIELDESAPO.
Django
Django is a critical level Python web system that
empowers quick advancement of secure and viable sites.
Worked by experienced engineers, Django deals with a
large part of the issue of web advancement, so you can
focus in on making your application without hoping to
reiterate an all around tackled issue.It is free and open
source, has a flourishing and dynamic local area,
extraordinary documentation, and numerous choices for
nothing and paid-for help
VI. Conclusion
In this venture, an examination to group Citrus Fruits and
Leaf Classification over static pictures it was created to
utilize profound learning procedures. This is a perplexing
issue that has previously been moved toward a few times
with various procedures. While great outcomes have been
accomplished utilizing highlight designing, this task zeroed
in on include realizing, which is one of DL guarantees.
While highlight designing isn't required, picture pre-
handling supports arrangement accuracy. Henceforth, it
decreases commotion on the information. These days,
Agriculture based AI Citrus natural products and leaf
incorporates is vigorously required. The arrangement
completely founded on highlight learning doesn't appear
to be close yet a direct result of a significant restriction.
Hence, Citrus organic products and leaf characterization
could be accomplished through profound learning
methods.
Future Work
Executing the Citrus Fruits and Leaf in a continuous
environment. To robotize this interaction by show the
forecast bring about web application or work area
application. To enhance the work to carry out in Artificial
Intelligence climate.
REFERENCES
[1] Asad Khattaki , Muhammad Usama Asghar2 , Ulfat
Batool2 , Muhammad Zubair Asghar 2 , Hayat Ullah2 ,
Mabrook al-rakhami 3 , (Member, IEEE), and Abdu gumaei
3 “Automatic Detection of Citrus Fruit and Leaves Diseases
Using Deep Neural Network Model” Received May 29,
2021, accepted July 2, 2021, date of publication July 13,
2021, date of current version August 18, 2021
[2] Stefania Barburiceanu , Serban Meza , (Member, IEEE),
Bogdan Orza , Raul Malutan, and Romulus Terebes,
(Member, IEEE), “Convolutional Neural Networks for

Texture Feature Extraction. Applications to Leaf Disease
Classification in Precision Agriculture” Received October
22, 2021, accepted November 19, 2021, date of publication
November 25, 2021, date of current version December 9,
2021.
[3] C. Senthilkumar, M. Kamarasan , “An Effective
Classification of Citrus Fruits Diseases using Adaptive
Gamma Correction with Deep Learning Model” , 2019.
[4] Mrs Rex Fiona1, Shreya Thomas2, Isabel Maria J3,
Hannah B4 , “Identification Of Ripe And Unripe Citrus
Fruits using Artificial Neural Network” , June 2019.
[5] Mohammed A. Alkahlout, Samy S. Abu-Naser, Azmi H.
Alsaqqa, Tanseem N. Abu-Jamie , “ Classification of Fruits
Using Deep Learning ” , Vol. 5 Issue 12, December – 2021.
[6] Sk Mahmudul Hassan 1 , Arnab Kumar Maji 1,*, Michał
Jasi ´nski 2,* , Zbigniew Leonowicz 2 and Elzbieta Jasi ´nska
˙ 3, “Identification of Plant-Leaf Diseases Using CNN and
Transfer-Learning Approach” , Received: 24 May 2021
Accepted: 8 June 2021 Published: 9 June 2021
[7] Ajit A.Bhosale* , “Detection of Sugar content in citrus
fruits by capacitance method “, 2016.
[8] Anisha Syal1, Divya Garg2, Shanu Sharma3 , “ A Survey
of Computer Vision Methods for Counting Fruits and Yield
Prediction “ , ISSN : 2319-7323 Vol. 2 No.06 Nov 2013.
[9] Prakanshu Srivastava1 , Kritika Mishra1 , Vibhav
Awasthi1 , Vivek Kumar Sahu1 and Mr. Pawan Kumar Pal2
,” Plant disease detection using Convolutional Neural
Network”, January 2021

Convolutional Neural Network for Leaf and citrus fruit disease identification using deep learning model

More Related Content

Similar to Convolutional Neural Network for Leaf and citrus fruit disease identification using deep learning model (20)

More from IRJET Journal (20)

Recently uploaded (20)

Convolutional Neural Network for Leaf and citrus fruit disease identification using deep learning model