International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249 – 8958, Volume-9 Issue-1, October 2019

Mulberry Leaf Disease Detection using Deep

Learning
D. Deva Hema, Sougata Dey, Krishabh, Anubhav Saha

Abstract: Disease diagnosis and classification in a mulberry Automatic detection of diseases by identifying their
plant using deep learning is an interesting technique which can symptoms of the plant leaves is easier and cheaper. Machine
be useful for farmers and researchers to identify and classify
vison and detection provides more accuracy compared
diseases. It helps to manage plant pathogens within fields
effectively and automatically at a minimal cost. Major mulberry traditional system. Deep learning is modern technique used
diseases usually express their symptoms on leaf area at the early for data analysis, image segmentation and image processing
stage of infection. Infections can be analysed and classified by technique in recent times. It can be applied in various
processing the image using a computer or machine using domain and now being used in agriculture as well.
different algorithms to interpret the information. This paper
Convolutional Neural Network (CNN) is used for image
gives us a brief knowledge of mulberry leaf diseases which is
used for automatic detection of disease. It presents in detail that processing. The early detection could help silk producers
the algorithm and techniques which are involved in and farmers to detect the disease which could be valuable
classification based on different criteria for image for plant growth as only fresh leaves are consumed by
segmentation. Our goal is to develop a more suitable deep silkworms [2,8]. The project done is to classify the type of
algorithm for our task. These convolutional layers are mostly
disease caused in the mulberry plan leaves. This would help
used for image processing. The system identifies and classify
mulberry leaf diseases effectively with complex scenarios from the farmer to take prerequisite measures steps for the
the affected areas using CNN. prevention of the disease from spreading more in the plant.
Mulberry leaf is very essential as it provides various
Keywords: Mulberry diseases, Convolutional neural
network, Leaf Spots, Powdery Mildew. products like silk, tea(very good for health). The leaf of the
mulberry plant is also used in the production of various
I. INTRODUCTION types of cosmetic products for both men and women.
Though mulberry cultivation is practiced in diverse climates
Agriculture is more than a feeding source in today’s world. and weather conditions, the major region of cultivation falls
Indian economy highly depends on productivity. The use of in the tropical zone of India covering the states of Andhra
pesticides and chemicals to increase productivity has very Pradesh, Karnataka and Tamil Nadu, which ranges to about
harmful effects on the soil, water and air. The plants such as 90% of the total growth. The states lying in sub topical
mulberry involve a high risk of crop failure. These are very region of India such as West Bengal, Himachal Pradesh and
expensive for production, so needs to be taken care of very some states of the north-east have major regions spared for
well. The right amount of fertilizers and pesticides should be mulberry cultivation. This proves many people (farmers) are
added to prevent harmful effects. It is generally cultivated in dependent on the production of cultivation of the mulberry
the Middle Eastern part of the world, Southern parts of plants. I hope our project if ever implemented in real life
Europe and the Indian subcontinent. Silkworm feeds only on will help the farmers to prevent their losses. Using the CNN
mulberry leaves. Only fresh mulberry leaves should be fed algorithm, we have performed the layer classification of the
as silkworms do not drink water, so all moisture is provided disease-induced leaf.
to them by leaves. The early detection is necessary to
prevent more infection, also bad will leaves will cause plant II. MULBERRY LEAF DISEASE
disease and the fruit production will be affected. Organic
farming is the solution for these problems which involves “Leaf Spots” (fungi – “Cercospora moricola”, “C.
pest control as well as disease control [1,14]. The classical missouriensis” and “Cercosporella spp”):
approach for disease detection is through naked eye 1)In the rainy season the leaf of the mulberry plants are
observation of the mulberry plants. It involves expert generally effected by fungi which can lead to defoliation of
the older trees.
observation and continuous monitoring. On large scale, team
2)Pre-treatment of the specimen should be done using
of experts are required, which costs very high for us. In antifungal pesticide.
many countries, farmers do not have proper facilities, also Symptoms:
they do not even have idea to contact the experts. 1)Formation of Irregular leaf spots or brown circular rings
are formed in the initial stage followed by spot enlargement,
Revised Manuscript Received on October 15, 2019 coalesce and formation of shot hole in the later stage.
D.Deva Hema, Assistant Professor in SRM Institute of Science and
Technology in Ramapuram, Chennai, Tamil Nadu, India.
2)Extremely affected leaves become pale yellow in colour
Sougata Dey, pursuing B. Tech degree in Computer Science and and fall off prematurely.
Engineering from SRM Institute of Science and Technology, Chennai
Krishabh is a third-year B.tech Computer science Engineering student
at SRM institute of science and technology, Chennai.
Anubhav Saha is currently pursuing B.Tech in the field of Computer
Science and engineering in SRM Institiute Of Science and Technology,
Ramapuram.

Retrieval Number: A1521109119/2019©BEIESP Published By:

Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.A1521.109119 3366 & Sciences Publication
 Mulberry Leaf Disease Detection using Deep Learning

III. RELATED WORKS

Some of the already present ideas on image processing and
leaf disease detection are taken for the reference purpose.
Vijay Singh/AK Mishra [1] presents generic algorithm for
Image segmentation using soft computing techniques. The
image of the leaf is taken and then converted into greyscale
image. Using texture feature including contrast, local
homogeneity, cluster shades are also used. For image
classification, SVM classifier and minimum distance
criterion is used.
Paper [2] presents disease classification using RCNN and
R-FCN methods. The data collection is done by taking
pictures and then image is preprocessed using image
annotation and augmentation. Image analysis is done using
multiple extractor. For experimentation dataset is divided
into testing, training and validation sets. System can
classify the diseases using algorithm.
Paper [3] presents the image segmentation including all
existing factors using the HSI color system, here H
component is used to segment spots and to reduce
illumination. The regions with disease spot segmented
using Sobel operator. Disease is finally graded by the
calculation of leaf spots.
Fig 1: Leaf Spot
Sanjay B. Patil et al. [4] presents thresholding methods. The
“Powdery mildew” :(fungi – “Phyllactinia corylea”
triangle and simple threshold techniques are used to
and “Uncinula geniculate”):
1)White patches occur on the leaf and which gradually gets segregate the affected region and leaf segment area
increased as it gets affected by Phyllactinia corylea. respectively. In the final step, the disease is classified using
Symptoms: quotient calculation of leaf and lesion area. According to
1)Initially, white powdered patches are observed on the this research, this method is fast and calculation of leaf
lower surface of leaves which eventually spread over the disease severity is accurate, here leaf area is calculated by
entire leaf surface. using threshold segmentation.
2)Later turn black to brown in color. Author [5] presents PNN algorithm to classify the diseases
3)Infected leaves turn yellow and fall off. using radial basis layer and competitive layer. The leaf
image is pre-processed, and greyscale conversion is done.
Network training is done using large number of datasets to
improve the results. The pieces of leaves are tested, some
sample got low accuracy. The system can recognise
different types of leaves.
In paper [6], author presents disease detection technique in
apple (Malus domestica) using k-mean clustering and Bayes
classifier. Texture segmentation is done by using co-
occurrence matrix and k-mean clustering. PCA analysis is
done and by using Bayes classifier system can detect
disease.
Paper [7] presents disease detection using k-means
clustering and multi SVM algorithm. Feature extraction and
statistical features by GLCM formula. Different statistical
features energy, sum entropy, covariance, information
measure, entropy contrast. The system can detect the
disease with less accuracy.
Author [8] describes about the fungus Myrothecium
verrucaria and its effects on mulberry leaves. Fugus grew
normally between 5°C and 35°C. The colony diameter was
found to be 50mm at optimum temperature after 6 days.
The author gives methods to protect leaves from infection.
Author [9] presents image thresholding techniques and its
performance evaluation. Point
Fig 2: Powdery Mildew dependent thresholding

Published By:
Retrieval Number: A1521109119/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.A1521.109119 3367 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-9 Issue-1, October 2019

techniques includes p-tile method, how the human brain recognizes an object in spite of its
mode method, Otsu method, Histogram concavity analysis, varying attributes from one another. Our brain has a
entropic method and minimum error method. Region complex layer of neurons ,each layer holds some
dependent techniques Include histogram transformation, information about the object and all the features of the
secondary level grey scale statistics, Deravi and pal method, object are extracted by the neurons and stored in our
local thresholding, multi thresholding methods has been memory, next time when we see the same object the brain
analysed. Ostu, Johannsen & Bille and moment-are better matches the stored features to recognize the object, but one
threshold selection methods region uniformity and shape can easily mistake it as a simple “IF-THEN” function, yes it
measure. Otsu method is better than other methods. is to some extent but it has an extra feature that gives it an
Paper [10] presents disease classification using plant texture edge over other algorithms that is Self-Learning, although it
features. First, colour transformation for RGB input image cannot match a human brain but still it can give it a tough
is generated. Then green pixel masking and removal is done competition .
using threshold value, finally classification is done using Image is processed using the Basic CNN to detect the
SVM classifier. diseases in leaves. The data training in our CNN model has
The author [11] presents image segmentation techniques to to satisfy following constraints:
do a comparative study between Edgebased, thresholding k- 1) There should be no missing values in our dataset.
means clustering and regionbased extraction. after image 2) The dataset must distinctly be divided into training and
enhancement, the extracted objects are compared with the testing sets, either the training or the testing set shouldn’t
mean weighted distance of the object with the original contain any irrelevant data out of our model domain in case
image. Correlation between objects are matched in two of an image dataset all the images must be of the same size,
images, it is done to measure the similarity between the one uneven distribution of image size in our dataset can
objects. PCA algorithm is used and the objects are aligned decrease the efficiency of our neural network.
to take account of different orientations. The paper shows 3) The images should be converted into black and white
us that k-mean clustering is better than other approaches. format before feeding it into the convolution layer because
The paper [12] presents fruit disease detection using color reading images in RGB would involve a 3-D numPy matrix
and texture by using artificial neural networks. The author which will reduce the execution time of our model by a
presents a system to detect and classify diseases using k- considerable amount.
means clustering. Multiple samples were taken and using an 4) Any kind of corrupted or blurred images should also be
algorithm, the system classifies the fruit and diseases. First, trimmed from the database before feeding it into the neural
image acquisition is done to remove the consequence of network .
hardware which was used to fabricate, image is segmented. Now we have learned the data pre-processing rules, let us
Finally, feature extraction is done and then the image is dive right into the working of the convolutional neural
matched with the training dataset for pattern classification. network.
The author [13] presents disease detection in groundnut
using Back Propagation Algorithm. Colour renovation is
done an RGB image is formed for colour generation and
description. RGB image is converted into HSV. Feature
extraction is done and then using algorithm disease
detection is done.
The author in [14] reference uses the process of converting
the carbon copy of the leaf mage into a negative image
.Then fragmenting the analysed carbon copy and then
removal of the components in the fragments.
All the above-mentioned papers have been thoroughly
studied. These papers gave a lot of description about the
chosen topic and the basic foundation concept for this paper
stands on the bits of references we observed.
Fig 3: CNN layers
IV. PROPOSED SYSTEM A. Convolution layer
The invention of the CNN in 1994 by Yann LeCun is what This layer involves scanning the whole image for patterns
propelled the field of Artificial Intelligence and Deep and formulating it in the form of a 3x3 matrix. This
learning to its former glory. The first neural network named convolved feature matrix of the image is known as Kernel.
LeNet5 had a very less validation accuracy of 42% since Each value in the kernel is known as weight vector.
then we have come a long way in this field. Nowadays
almost every giant technology firms rely on CNN for more
efficient performance.
The idea to detect diseases in mulberry leaf incorporates the
use of CNN before we dive into the “functionality and
working of CNN” concept, we must have a basic idea on

Retrieval Number: A1521109119/2019©BEIESP Published By:

Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.A1521.109119 3368 & Sciences Publication
 Mulberry Leaf Disease Detection using Deep Learning

V. SYSTEM ARCHITECTURE

Fig 4: Convolution layer

B. Pooling layer
After the convolution comes to the pooling here the image
matrix is broken down into the sets of 4 rectangular
segments which are non-overlapping. There are two types
of pooling, Max pooling and average pooling. Max pooling
gives the maximum value in the relative matrix region
which is taken. Average pooling gives the average value in
the relative matrix region. The main advantage of the
pooling layer is that it increases computer performance and
decreases over-fitting chances.

Fig 7: Proposed System Architecture

VI. GRAPHICAL OUTPUT

Fig 5: Pooling layer

C. Activation layer
It the part of the Convolutional Neural Networks where the
values are Normalized that is, they are fitted in a certain
range. The used convolutional function is ReLU which
allows only the positive values and then rejects the
negative values. It is the function of low computational
cost.

Fig 8: Performance analysis 1

Fig 6: ReLU function Fig 9: Performance analysis 2

D. Fully connected layer
Here the features are compared with the features of the test
image and associate similar features with the specified
label. Generally, labels are encoded in the form of numbers
for the computational ease, they will be later converted into
their respective strings.

Published By:
Retrieval Number: A1521109119/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.A1521.109119 3369 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-9 Issue-1, October 2019

Republic where more than 60 percent of the GDP is

dependent on Agriculture. In some countries, the
production of silk contributes to about 20% of the overall
GDP. Now the only food that the silkworm can feed on is
the mulberry leaves, they don’t have any kind of other
diets. In these countries, any outbreak of disease among the
crops can be catastrophic and will result in a heavy loss of
lives and capital. Here the Convolutional Neural Network
System predicts the disease by analysing the first appearing
symptoms, it spares the farmers from a lot of effort. CNN
can also be used to predict the Cure of the Disease and The
proportion of the Fertilizers required for the Cure by
combining some logistic regression techniques with it, It
Fig 10: Performance analysis 3 can also inform the farmers nearest locations where he/she
could get the required fertilizers/pesticides and also their
Authentic prices so that they cannot be tricked by the some
greedy shopkeepers. Using Algorithm across various
platforms including mobile apps can help it spread easily to
the masses, Agriculture experts can also issue a question
forum where the farmers can post their issues which will be
later resolved by the agricultural experts.

REFERENCES
1. Vijai Singh, A.K Misra (2017). “Detection of plant leaf diseases using
Fig 11: Performance analysis 4 image segmentation and soft computing techniques. Information
Processing in Agriculture”, 4(1), 41–49.
2. M. Akila, P. Deepan (2018). “Detection and Classification of Plant
VII. CONCLUSION Leaf Diseases by using Deep Learning Algorithm”. ICONNECT –
2018 (VOLUME 6 – ISSUE 07)
The project done is to classify the type of disease caused by 3. Weizheng, S., Yachun, W., Zhanliang, C., & Hongda, W. (2008).
the mulberry plant. This would help the farmer to take prior “Grading Method of Leaf Spot Disease Based on Image Processing”.
steps for the prevention of the disease from spreading more 2008 International Conference on Computer Science and Software
in the plant. Mulberry leaf is very essential as it provides Engineering.
4. Sanjay B. Patil et al. “LEAF DISEASE SEVERITY
various products like silk, tea (very good for health). The MEASUREMENT USING IMAGE PROCESSING”. International
leaf of the mulberry plant is also used in the production of Journal of Engineering and Technology. Vol.3 (5), 2011, 297-301
various types of cosmetic products for both men and women 5. Wu, S. G., Bao, F. S., Xu, E. Y., Wang, Y.-X., Chang, Y.-F., &
Though mulberry cultivation is practiced in diverse climates Xiang, Q.-L. (2007). “A Leaf Recognition Algorithm for Plant
Classification Using Probabilistic Neural Network”. 2007 IEEE
and weather conditions, the major region of cultivation falls International Symposium on Signal Processing and Information
in the tropical zone of India covering the states of Andhra Technology.
Pradesh, Karnataka and Tamil Nadu, which ranges to about 6. Sabah Bashir, Navdeep Sharma. “Remote Area Plant Disease
90% of the total growth. The states lying in sub topical Detection Using Image Processing”. IOSR Journal of Electronics and
Communication Engineering. ISSN: 2278-2834 Volume 2, Issue 6
region of India such as West Bengal, Himachal Pradesh and 2012, PP 31-34
some states of the north-east have major regions spared for 7. Sandesh Raut, Amit Fulsunge (2017). “Plant Disease Detection in
mulberry cultivation. This proves many people (farmers) are Image Processing Using MATLAB”. Vol. 6, Issue 6, June 2017.
dependent on the production of cultivation of the mulberry IJIRSET
8. Murakami, R., Kobayashi, T., & Takahashi, K. (2005). “Myrothecium
plants. I hope our project if ever implemented in real life leaf spot of mulberry caused by Myrothecium verrucaria. Journal of
will help the farmers to prevent their losses. Using the CNN General Plant Pathology”, 71(2), 153–155. doi:10.1007/s10327-
algorithm, we have performed the layer classification of the 0040178-8
disease-induced leaf. 9. Sahoo P K, Soltani S, Wong A K C, et al. “A survey of thresholding
techniques [J]. Computer Vision, Graphics, and Image Processing”.
1988, 41(2)
VIII. FUTURE WORK 10. S. Arivazhagan, R. Newlin Shebiah*, S. Ananthi, S. Vishnu Varthini.
“Detection of unhealthy region of plant leaves and classification of
The convolutional neural network is a category of deep plant leaf diseases using texture features”. CGIR 2015 vol.15.
learning strategies that have become dominant in various 11. Sapna Varshney, S., Rajpal, N., & Purwar, R. (2009). “Comparative
computer vision tasks in recent times and is attracting study of image segmentation techniques and object matching using
interest across a widespread diversity of domains, segmentation”. 2009 Proceeding of International Conference on
Methods and Models in Computer Science (ICM2CS).
including crop cultivation agriculture. Looking at its 12. Ashwini Awate, Damini Deshmankar, Prof. Samadhan Sonavane
current growth and its widespread uses it can be deduced “Fruit Disease Detection using Color, Texture Analysis and ANN”
that soon enough to find neural networks controlling some IEEE International Conference on Green Computing and Internet of
major aspects of our day to day life. There are many Things (ICGCIoT), pp. 978-1-4673-7910, 2015.
countries in the word like Guinea-Bissau, Central African

Retrieval Number: A1521109119/2019©BEIESP Published By:

Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.A1521.109119 3370 & Sciences Publication
 Mulberry Leaf Disease Detection using Deep Learning

13. Ramakrishnan.M and Sahaya Anselin Nisha.A “Groundnut Leaf Anubhav Saha is currently pursuing B.Tech in
Disease Detection and Classification by using Back Probagation the field of Computer Science and engineering
Algorithm” IEEE ICCSP conference, pp. 978-1-4 799-8081-9/15, in SRM Institiute Of Science and Technology,
2015 . Ramapuram. completed class 12 from India
14. “Identification of diseases in plant parts using image processing” by International School in 2017.Fields of Interest
S.S. Saranya, Nalluri Chandra Kiran,Komma ,Jyotheeswar Reddy. include NLP, Machine Learning and Qunatum
Computing.
15. A. Meunkaewjinda, P.kumsawat, K.Attakitmongcol and Asrikaew.
“Grape leaf disease detection from color imagery using hybrid
intelligent system” proceedings of ECTi-CON 2008.
16. Prof. Sanjay B. Dhaygude, Nitin P.Kumbhar. “Agricultural plant leaf
disease detection using image processing”. International Journal of
Advanced Research in Electrical, Electronics and Instrumentation
Engineering Vol. 2, Issue 1, January 2013.
17. Otsu N. “A threshold selection method from graylevel histogram” [J].
IEEE Transations on System, Man, and 1979(9): 652-655
18. Cao YuePing, Yang Qingkai. “Research on the standard anti-flu of
Soybean gray leaf spot disease”. Soybean Science. 2002, 21(2):113-
116.
19. Jayamala K. Patil , Raj Kumar “Advances in Image Processing for
Detection of Plant Disease” Journal of Advanced Bioinformatics
Applications and Research ISSN 0976-2604 Vol. 2, Issue 2, June-
2011, pp 135- 141.
20. Savita N. Ghaiwat, Parul Arora. “Detection and Classification of
Plant Leaf Diseases Using Image Processing Techniques”: A Review.
International Journal of Recent Advances in Engineering &
Technology (IJRAET) ISSN (Online): 2347 - 2812, Volume-2, Issue -
3, 2014.
21. Prakash M. Mainkar, Shreekant Ghorpade , Mayur Adawadkar. “Plant
Leaf Disease Detection and Classification Using Image Processing
Techniques”.
22. Rao, VM. Inclusive growth in India – “Agriculture, poverty and
human development. Indian Journal of Agricultural Economics”. Vol.
63, Iss. 2, (Apr-Jun 2008), 276-279.

AUTHORS PROFILE
D.Deva Hema received the M.E Degree in Anna
University in 2007 and pursuing Ph.D. in
Satyabhama Institute of Science and Technology.
She is currently Assistant Professor in SRM
Institute of Science and Technology in
Ramapuram, Chennai, Tamil Nadu, India. Her
research interests include the area of Artificial
Intelligence and Machine Learning including
crash prediction and occupant protection during
vehicular crashes. She is currently focusing on crash prediction for
intelligent transportation systems, injury mechanisms during rollover
crashes, and the protection of pedestrians and drivers.

Sougata Dey is currently pursuing B. Tech

degree in Computer Science and Engineering
from SRM Institute of Science and Technology,
Chennai and will be graduated in 2021. He has
strong interest in Machine Learning, Natural
language processing, Artificial Intelligence and
Data Science. He wants to do his master degree
with specialization in Machine Learning. He
aspires to become a data scientist. His main goal
is to put down his knowledge in bringing up
something productive for the society.

Krishabh is a third-year B.tech Computer

science Engineering student at SRM institute of
science and technology, Chennai. He received
his 12th from G.A inter school, Hajipur. He has
strong interest in Artificial intelligence and
machine learning. He also likes developing new
software and apps.

Published By:
Retrieval Number: A1521109119/2019©BEIESP Blue Eyes Intelligence Engineering
DOI: 10.35940/ijeat.A1521.109119 3371 & Sciences Publication