Convolutional Neural Networks For The Detection of Multiclass Plant Diseases

Volume 7, Issue 5, May – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Convolutional Neural Networks for the Detection of

Multiclass Plant Diseases
Dona Joby, Christeena Shaju, Fasna T A., Avani Das
Department of Computer Science and Engineering
Sahrdaya College of Engineering and Technology
Abstract:- Most industrialised countries' economies are proposed a number of strategies to address the issues
based on agriculture. Crop production is one of the most mentioned above. For the categorization of plant diseases,
influential factors in a country's domestic market many types of feature sets may be employed in machine
scenario. Agricultural output is also a crucial component learning. Traditional methods and deep-learning-based
of every country's economic development. Agriculture is characteristics are the most seen features among them. The
critical because it offers raw materials, work, and food to most promising way for automatically learning decisive and
a diverse population. Overuse of chemical fertilisers, discriminative characteristics is to use deep-learning-based
pollution of water supplies with chemicals, irregular algorithms, specifically CNNs. Deep learning (DL) is made
rainfall patterns, shifting soil fertility, and other factors up of several convolutional layers that reflect data learning
are among them. Apart from these challenges, disease- characteristics. Using a deep-learning algorithm, plant
related loss of a significant section of output is one of the diseases may be detected. Deep learning has certain
most prominent roadblocks across the world. The disadvantages, such as the fact that it takes a lot of data to
presence of illnesses in the grown plants decreases a train the network. Performance suffers if the supplied dataset
major share of the yield after delivering efficient lacks sufficient pictures.
resources to the fields. As a result, scientists have been
working on a new project. As a result, scientists are II. LITERATURE SURVEY
focused their efforts on developing effective ways for
detecting illness in plants. Plant diseases are a major Manual procedures are used to validate and manage
problem for small-scale farmers because they disrupt the plant disease in the majority of cases. Unaided eye
food supply. To provide efficient processes for diagnosis perception is one such important approach. In any event, this
and avoidance of destruction, it is necessary to identify strategy necessitates continual observation of the region by
the kind of plant disease existing as soon as possible. someone with better knowledge of the plants and illnesses
Significant progress has been achieved in discovering that affect them. Likewise, the expert must be available in a
plant diseases that impact a range of crops in different timely manner, or else it may result in loss. The presence of a
regions of the world in recent years. Image capture, disease on a plant can also be determined by research centre
preprocessing, and segmentation are all steps in the testing. The harvest's ability to fight back is being harmed by
process of detecting plant diseases. It's additionally this manual testing procedure.
enhanced by a number of feature extraction and III. DEEP LEARNING ALGORITHMS
classification methods.
CNN Algorithm: A convolution neural network (CNN)
Keywords:- Plant disease, VGG16, InceptionV3, Resnet50, is a specific type of artificial neural network that uses
Hybrid model. perceptrons, an AI unit computation, for supervised learning
I. INTRODUCTION and information inquiry. CNNs are used in image
preprocessing, speech processing, and a variety of other
In this world, detection of plant diseases from diseased cognitive tasks. A ConvNet is a name for a convolutional
plant leaves is a significant development in agriculture. neural network.
Furthermore, early and precise agricultural disease
identification enhances crop output and quality. Because of VGG16 Model: The VGG16 architecture is built on a
the great diversity of agricultural products, even farmers and convolutional neural network (CNN). It is widely
pathologists may struggle to identify plant diseases from sick acknowledged to be one of the most effective vision model
leaves. However, in poor nations' rural regions, eye designs to date. VGG16 is remarkable for having 3x3 filter
inspection remains the most prevalent technique of illness convolution layers with a stride 1 and always using the same
detection. It is also necessary to have expert monitoring. padding and maxpool layer of 2x2 filter stride 2. Throughout
Farmers in rural places must drive long distances to meet the design, the convolution and max pool layers are
with specialists, wasting both time and money. In finding and positioned in the same order. At last, there are two FC
diagnosing plant diseases, farmers benefit from the high (completely connected layers) and a softmax layer as
throughput and precision of automated computer systems. output.The 16 in VGG16 alludes to the fact that there are 16
Farmers and agronomists benefit from the high throughput layers with different weights. This network is much larger,
and precision of automated computational systems for with over lakhs(approximation) of parameters. Two
detecting and diagnosing plant diseases. Researchers have additional dense layers added along the VGG16 model.
IJISRT22MAY2052 www.ijisrt.com 1380

ISSN No:-2456-2165
information sources and, if not, the accuracy of its
Resnet50 Model: ResNet-50 is a 50-layer deep estimations. Because it trains, this aids in directing the
convolutional neural network. You may use the ImageNet neurological system to reinforce the proper beliefs. During
database to load a pretrained version of the network that has training, this is always the final layer.
been trained on over a million photos. This network has
ability to categorize various images into hundreds of IV. ATTRIBUTES AND DATASET
different item categories, including books, birds, and a broad
variety of animals. As a result, the network has picked up a This dataset was reconstructed from the original dataset
variety of rich feature representations for a variety of photos. using offline augmentation. This github project contains the
The network's picture input size is 224 by× 224 pixels. The original dataset. This collection contains around 87,500 rgb
Resnet50 model here has two more additional dense layers photos of healthy and sick crop leaves, which are divided
with Relu activation functions. into 38 classifications. The complete dataset is divided into
an 80/20 training and validation set, with the directory
InceptionV3 Model: Inception-v3 is a 48-layer deep structure preserved.
convolutional neural network model that has been pre-
trained. It's a version of the network that's already been V. CONCLUSION
trained on millions of photos from the ImageNet collection.
It's the third iteration of Google's Inception CNN model, The use of monitoring checking and executive
which was first proposed during the ImageNet Recognition frameworks is rising in demand with the advancement of
Challenge. Convolutional Neural Networks employ Inception technology, according to this article. In the world of
Modules to provide more efficient computation and deeper gardening. The majority of crop loss occurs as a result of a
networks by reducing dimensionality using stacked 11 long distance, the spread of illness The model's accuracy is
convolutions. Two additional dense layers added along the 87.47 percent. We have compared three different models of
InceptionV3 model. CNN models in which two dense layers are added to each
VGG16, ResNet50, and InceptionV3 models and found that
Layers of CNN are: VGG16 models provides higher accuracy and hence leads to
 Convolution Layer- CNNs are a type of Neural Network find better CNN model for plant diseas detection.
that has shown to be incredibly effective in a variety of
Future works in the project can be done by developing a
situations. Recognize and classify images, for example.
hybrid model which can further provide much higher
CNNs are multilayered feed-forward neural networks.
accuracy. AutoML can be employed which will provide
CNNs are made up of filters, kernels, or neurons with
more accuracy.
learnable loads, inclinations, and parameters. Each filter
takes a few input components, convolutions them, and
alternates between them.
 Pooling Layer- The pooling layer reduces the complexity
of each activation map while maintaining the most
important information. The images are divided into a large
number of non-covering square forms.A non-linear
procedure, such as average or maximum, is used to check
each zone. This layer, which is frequently placed between
convolutional layers, achieves superior speculation, faster
assembly, and is powerful to interpretation and distortion.
 Activation Layer- The activation layer regulates how
signals flow from one layer to the next, simulating how
neurons in our brain are terminated. More neurons would
be activated by output signals that are strongly linked to
previous references, allowing indications to propagate more
effectively for identification. CNN is equipped with a wide
range of complicated enactment capacities for
demonstrating signal propagation, the most basic of which
is the Rectified linear measure (ReLU), which is favoured
for its faster processing speed.
 Fully Connected- The system's last levels are fully
connected, meaning that neurons from previous layers are
linked to neurons in subsequent layers. This is similar to
high-level thinking, in which every possible path from
input to output is considered.
 Loss - During the training of the neural network, there is an Fig. 1
additional layer known as the misfortune layer. This layer
assesses the neural network's ability to distinguish

ISSN No:-2456-2165
Function," 2020 International Conference on
Electronics and Sustainable Communication Systems
(ICESC), 2020, pp. 564-569, doi:
10.1109/ICESC48915.2020.9155815
[9.] S. S. Chouhan, U. P. Singh, and S. Jain, ‘‘Web
facilitated anthracnose disease segmentation from the
leaf of mango tree using radial basis function (RBF)
neural network,’’ Wireless Pers. Commun., vol. 113,
no. 2, pp. 1279–1296, Jul. 2020, doi:10.1007/s11277-
020-07279-1
[10.] K. S. Patle, R. Saini, A. Kumar, S. G. Surya, V. S.
Palaparthy and K. N. Salama, "IoT Enabled, Leaf
Wetness Sensor on the Flexible Substrates for In-Situ
Plant Disease Management," in IEEE Sensors Journal,
vol. 21, no. 17, pp. 19481-19491, 1 Sept.1, 2021, doi:
10.1109/JSEN.2021.3089722.
[11.] Jaidhar C. D , Sunil C.K and Nagamma Patil S ,”
Cardamom Plant Disease Detection Approach Using
EfficientNetV2”, Volume 10, December 27,
Fig. 2 doi:10.1109/ACCESS.2021.3138920.
REFERENCES
[1.] Zongshuai Liu, Xuyu Xiang, Jiaohua Qin, Yun Tan,

Qin Zhang and Neal N. Xiong R. Nicole, “Image
Recognition of Citrus Diseases Based on Deep
Learning,” CMC, 2021, vol.66, no.1,
doi:10.32604/cmc.2020.012165.
[2.] Sophia L. Sanga , Dina Machuve and Kennedy
Jomanga,” Mobile-based Deep Learning Models for
Banana Disease Detection,” Engineering, Technology
& Applied Science Research, Vol. 10, No. 3, 2020,
5674-5677
[3.] E. L. da Rocha, L. Rodrigues, and J. F. Mari, ‘‘Maize
leaf disease classification using convolutional neural
networks and hyperparameter optimization,’’ in Proc.
Anais do XVI Workshop Visão Computacional. (SBC),
2020, pp. 104–110.
[4.] Hassan Amin, Ashraf Darwish, Aboul Ella Hassanien
and Mona Soliman,” End-to-End Deep Learning Model
for Corn Leaf Disease Classification,”Volume 10,2022,
10.1109/ACCESS.2022.3159678
[5.] J. Zhu, A. Wu, X. Wang, and H. Zhang, ‘‘Identification
of grape diseases using image analysis and BP neural
networks,’’ Multimedia Tools Appl., vol. 79, nos. 21–
22, pp. 14539–14551, Jun. 2020, 10.1007/s11042- 018-
7092-0.S. S.
[6.] F. Saeed, M. A. Khan, M. Sharif, M. Mittal, L. M.
Goyal, and S. Roy, ‘‘Deep neural network features
fusion and selection based on PLS regression with an
application for crops diseases classification,’’ Appl.
Soft Comput., vol. 103, May 2021, Art. no. 107164.
[7.] S. H. Lee, H. Goëau, P. Bonnet, and A. Joly, ‘‘New
perspectives onplant disease characterization based on
deep learning,’’ Comput. Electron.Agricult., vol. 170,
Mar. 2020, Art. no. 105220.
[8.] S. Y. Yadhav, T. Senthilkumar, S. Jayanthy and J. J. A.
Kovilpillai, "Plant Disease Detection and Classification
using CNN Model with Optimized Activation

Convolutional Neural Networks For The Detection of Multiclass Plant Diseases

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Convolutional Neural Networks For The Detection of Multiclass Plant Diseases

Uploaded by

Copyright:

Available Formats

Volume 7, Issue 5, May – 2022 International Journal of Innovative Science and Research Technology

Convolutional Neural Networks for the Detection of

IJISRT22MAY2052 www.ijisrt.com 1380

IJISRT22MAY2052 www.ijisrt.com 1381

[1.] Zongshuai Liu, Xuyu Xiang, Jiaohua Qin, Yun Tan,

IJISRT22MAY2052 www.ijisrt.com 1382

You might also like