IJSRD - International Journal for Scientific Research & Development| Vol.

11, Issue 5, 2023 | ISSN (online): 2321-0613

Agriculture Plant Disease Investigation and Image Classification with

Artificial Neural Network Using ML
Mr. M. Manikandan1 R.Deepak2
1
Professor 2PG Scholar
1,2
Department of Computer Science Engineering
1,2
Vidhyaa Vikas College of Engineering and Technology, Tiruchengode, India
Abstract — The position of agrarian product is pivotal to a until recent and lower dataset won't give accurate bracket. So,
nation’s profitable growth. The biggest handicap to the Plant Village design was initiated to profit growers with early
product and quality of food, however, is factory complaint. complaint discovery and started collecting thousands of
Beforehand discovery of factory conditions is essential for images of all types of crops. This chapter focuses on
maintaining global health and weal. The standard system of automatic factory complaint discovery as a content of
opinion entails a pathologist visiting the position and visually discussion.
assessing each factory. still, due to lower delicacy and limited The study demonstrates the specialized feasibility of
availability of mortal coffers, homemade examination for deep literacy to enable automatic tomato factory complaint
colorful factory conditions is limited. To address these discovery through tomato splint images. The dataset
problems, it's necessary to develop automated styles that can correspond of both diseased and healthy splint images. In this
snappily identify and classify a wide range of factory chapter, deep complication neural network is used to classify
conditions. The presence of low- intensity information in the the factory splint as healthy or diseased. colorful conditions
image background and focus, the extreme colour similarity of tomato splint is linked liketomato_mosaic, lateblight,
between healthy and diseased factory areas, the presence of unheroic twisted, septoria Leaf Spot, healthy, bacterial spot
noise in the samples, and changes in the position, with total 21071 images. These conditions can be linked at
chrominance, structure, and size of factory leaves make it original stage using CNN and growers can use any infection
delicate to directly identify and classify factory conditions. control tools to stop spreading complaint to other shops and
We've developed a dependable factory complaint bracket also break pest problems while minimizing the pitfalls of
system using an InceptionV3 Architecture to address the mortal and terrain. Some of the challenges linked are as
forenamed issues. In this exploration, we suggested a deep follows.
literacy strategy grounded on InceptionV3 Architecture to  Quality of the splint image
identify splint conditions in a variety of shops. Chancing the  Larger dataset
factory complaint and its bracket is our end. The substantiated  Image denoising
dataset is taken from the well- known public source kaggle.  Segmenting exact spot of complaint in splint
The dataset consists of 70,295 Factory images of Apple,  unyoking the training and testing samples from original
Blueberry, Cherry, Corn (sludge), Grape, Orange, Peach, images
Pepper bell, Potato, jeer, Soybean, Strawberry and Tomato.  point birth like color, size, texture and shape from image
The suggested system has the capacity to handle complex
 Feting different type of complaint from factory leaves
situations from a shops area and can successfully identify
Leaf conditions are trouble to the crop product and
colorful forms of conditions. planter’s frugality reduces immensely with increase in spread
Keywords: Agriculture Plant Disease, Image Classification, of crop complaint. Conventional styles of detecting crop
Artificial Neural Network, ML conditions bear a great deal of knowledge and moxie. Further,
these styles can be precious, time consuming, and indeed
I. INTRODUCTION ineffective in some cases. These challenges can be answered
changeable climate change, without any proper irrigation using deep literacy ways which are having further implicit to
fashion, lack of knowledge in using ultramodern technologies identify colorful conditions. This motivates the work to
and use of fungicides have caused an imbalance in husbandry concentrate substantially on splint complaint discovery and
which has caused indecorous civilization and unhealthy crops classify the complaint at early stage with great delicacy and
thereby hanging food security. Factory pathogens will lead to help growers from huge loss. The study uses tomato splint for
crop loss. Due to creatures, weeds, pathogens it's estimated to analysis as tomato is considered most profitable crop because
reduce agrarian product between 20 to 40 encyclopedically. it's grown four times in a time. Factory complaint is relatively
Agriculture is one the main sources of income in India it natural because they cannot repel light intensity, high
struggles to support fleetly growing population. moisture, temperatureetc., it affects factory saturation, fruit
It's estimated that in India smallholder growers will color and fruit quality and temperature below 10 degree
induce around 80 percent of agrarian product and due to pests Celsius and above 38 degree Celsius affects factory towel and
and factory complaint there's further than 50 percent of yield factory.
loss. Factory conditions reduce the quality and product of Due to all these environmental conditions shops are
food crops. To overcome this we need large vindicated susceptible to conditions like fungus, bacteria,
dataset of images which includes healthy and diseased splint contagionsetc., Early discovery of factory complaint can help
images of all type of crop shops. Using this dataset develop in adding yield by precluding from complaint. CNN improves
an accurate image classifier which classifies the type of the bracket delicacy in numerous fields, including husbandry.
complaint at original stage. similar dataset wasn't available

All rights reserved by www.ijsrd.com 17

 Agriculture Plant Disease Investigation and Image Classification with Artificial Neural Network Using ML
(IJSRD/Vol. 11/Issue 5/2023/006)

So, deep complication neural network is used for B. Objects

demonstration. There are three objects of the proposed methodology
The model proposed mongrel deep literacy model, 1) To develop a prototype for a factory complaint discovery
which consists of transfer literacy, attention model, and system.
powerhouse operation, is proposed along with a CNN- 2) To apply image processing ways to identify the
grounded VGG16 classifier to descry and classify conditions complaint pattern.
associated with shops. Pretrained imageNet weights were 3) Use machine literacy algorithms to prognosticate
used for image preprocessing. Pre-trained images are trained complaint.
on millions of images with 1000 different orders, which 4) Use transfer literacy ways to prognosticate complaint.
reduce the time for training new images. Further, all images
were 3- dimensional, i.e., they were measured in terms of C. Compass
height, range, and channels( RGB). The proposed methodology is used for the precise discovery
of complaint in crops. Which can give controlled fertilization
II. LITERATURE REVIEW to growers. Accurate identification of complaint also helps
According to Ehsan Kianiet.al, 2017( 16) image segmentation growers to identify the infection and do fairly controlled
done with the help of colors i.e. color image segmentation fertilization to avoid any unborn crop failures.
ways helps to more understand and break the problem. One D. Flow chart for the proposal model
can first find out the three- color image factors of an image
which are Red, Green, and Blue factors. The red and Green
factors help to identify the unheroic factors of the image
which is generally marked as an infected part. Fuzzy sense is
a good fashion to break a complaint bracket problem.
Vijai Singhet.al( 2017) An advancement of
inheritable algorithm is proposed by the author named
minimal distance algorithm to find the infected factory part
of the factory that's to perform image segmentation. After the
image segmentation step the author has checked the delicacy
of the algorithm with other bracket algorithms like k mean
clustering and SVM( 17).
Konstantinos 2018( 18) In this paper author has used
a complication neural network fashion to identify colorful
factory conditions. A detailed study has been done by the
author. Images of colorful factory leaves are taken which
includes both the infected leaves images and healthy leaves
images and also the author has classified it in colorful classes
and all CNN infrastructures gave further than 97 delicacy.
The CNN infrastructures include AlexNet, AlexNetOWTBn, Fig. 3.4: Flow chart for the proposal model
GoogLeNet, Overfeat, VGG. E. Module Description
Kamlesh Golhaniet.al, 2018 In this paper author has
Loading Dataset( RGB Images) The first stage of plant
done a detailed review of colorful deep literacy algorithms
complaint identification and type system is loading dataset.
along with their advantages and disadvantages also their
This dataset contains high- quality RGB( red, green, blue)
optimization ways. A comparison has also been made for
plant images which include healthy plant images and
these ways about the affiliated work.( 19)
complaint plant images.
Pre-Processing The alternate stage of the plant
III. METHODOLOGY complaint discovery and type system ispre- processing. In
A. Problem Statement this stage remove noisy and inconsistent data from the dataset
The proposed exploration work is carried on three crops and extract only useful information. This system includes
Potato, tomato, and pepper. In a check, 61.33 of growers some ways like image resizing, image smoothing, image
cultivating potatoes reported bright as one of the major enhancement, etc.
reasons for their crop failure. According to The Hindu, in the Segmentation The third stage of the plant illness
time 2020, Maharashtra faced a huge extremity for tomato identification and type frame is segmentation. In this process
cropsapprox. 60 of crops failed due to a contagion. India has partitioning image into various corridor like converting RGB(
a 40 share of total pepper product encyclopedically. Pepper red, green, blue) to HSV( tincture, achromatism, value),
has also numerous herbal benefits in throat infections. India converting RGB( red, green, blue) to LAB( radiance, green/
is the 2nd largest patron of tomatoes and potatoes followed red, blue/ pusillanimous) and other ways. The segmentation
by China which is the largest patron of these crops. perfecting should be possible by exercising different procedures like
fertilization and automating the system for complaint Otsu' strategies, K- means clustering. point birth The fourth
discovery can ameliorate the crop product of our country. stage of the plant complaint identification and type approach
is point birth. point birth plays a significant part in image
type. point birth can be employed in numerous operations. In-

All rights reserved by www.ijsrd.com 18

 Agriculture Plant Disease Investigation and Image Classification with Artificial Neural Network Using ML
(IJSRD/Vol. 11/Issue 5/2023/006)

plant complaint type uses multitudinous features like color, employed for the double image segmentation and infected
texture, morphology, and edges, etc. This approach has region identification this was done with the help of preparing
tracked down that morphological issues give preferable issues an image mask. The Gray LevelCo-occurrence Matrix is the
over different features. It tends to be relating the infected main tool that implements the generalities learned from
plant flake of the type plant image. We can likewise liberate uprooted features. employed for point birth, &multi-class
unique shops and unique conditions from the dataset. support vector machine( SVM) methodology was employed
Bracket The fifth stage of the plant infection for the bracket of potato leaves. The model deduced is
recognition and type system is the type fashion. This type estimated using certain evaluation criteria perfection, recall,
stage performs any of the machine knowledge or deep F1- score, and delicacy.
knowledge ways for classifying the various conditions in
shops.
Discovery and Bracket of Plant Disease This is the
final outgrowth stage of our proposed system. In this stage,
stoners can identify and characterize the plant infection.
F. Building the Model Architecture Table 1: Evaluation Metrics of Potato Disease analysis using
ResNets Architecture solves one of the notorious problems SVM
known as sinking slants. In ResNets, unlike in traditional
neural networks, each caste feeds into the coming caste, we
use a network with residual blocks, each caste feeds into the
coming caste and directly into the layers about 2 – 3 hops
down, to avoidover- fitting( a situation when evidence loss Table 2: Details of evaluation metrics obtained for the
stop abating at a point and also keeps adding while training transfer learning models implemented
loss still decreases). This also helps in preventing sinking
grade problem and allow us to train deep neural networks. ”

IV. SAMPLE OUTPUTS

Fig. 5: Comparison of evaluation metrics across the models

implemented

VI. CONCLUSION
Fig. 4.1: Upload image The machine learning algorithm used in proposed work is
SVM. SVM gave good result in when the discovery orders
were less. As the no of complaint orders increased it failed to
achieve the delicacy. Transfer literacy is the current effective
exploration for carrying the better performance of the models
with a minimum and briskly training phase. It proved
veritably true with the proposed frame. The proposed frame
suitable to attain better delicacy with all the three models
similar as VGG- 16, ResNet- 50, and ResNet- 50 v2, yet
(1) Input Interface (2) Interface output ResNet- 50 grounded transfer literacy model a bit more
Fig. 4.2: Model loss and Accuracy effective when compared to the other models. The proposed
frame effective with themulti-class bracket of colorful
V. RESULTS AND DISCUSSION conditions along with healthy leaves that include crops of
Disease discovery using ML At first machine literacy pepper, potato, and tomato.
algorithm was enforced to descry factory complaint. This was
done in two phases.I.) enforcing ML algorithm on potato REFERENCES
dataset. II) enforcing ML algorithm on the entire dataset. [1] Sardogan, M., Tuncer, A., and Ozen, Y.: Plant Leaf
A. Enforcing ML Algorithm on Potato Dataset. Disease Detection and Classification Based on CNN
with the LVQ Algorithm. In: 3rd Int. Conf. Comput. Sci.
The dataset which is considered in the proposed work is an
Eng. (2018) 382–385
openly penetrated dataset & it was aimlessly divided into the
[2] Wallelign, S., Polceanu, M., and Buche, C.: Soybean
training dataset consists of 1820 images and the testing
plant disease identification using a convolutional neural
dataset consists of 780 images. The otsu algorithm was

All rights reserved by www.ijsrd.com 19

 Agriculture Plant Disease Investigation and Image Classification with Artificial Neural Network Using ML
(IJSRD/Vol. 11/Issue 5/2023/006)

network. In: Proc. 31st Int. Florida Artif. Intell. Res. Soc.
Conf. FLAIRS 2018 (2018), 146–151
[3] Sladojevic, S., Arsenovic, M., Anderla, A., Culibrk, D.,
and Stefanovic, D.: Deep Neural Networks Based
Recognition of Plant Diseases by Leaf Image
Classification. Comput. Intell. Neurosci. 2016(2016)
[4] Fuentes, A., Yoon, S., Kim, S. C., and Park, D. S.: A
robust deep-learning-based detector for real-time tomato
plant diseases and pests recognition. Sensors
(Switzerland) 17 (2017).
[5] Arivazhagan, S. and Ligi, S. V.: Mango Leaf Diseases
Identification Using Convolutional Neural Network. Int.
J. Pure Appl. Math., 120(2018) 11067–11079
[6] Oppenheim, D. and Shani, G.: Potato Disease
Classification Using Convolution Neural Networks.
Adv. Anim. Biosci., 8 (2017), 244–249
[7] Barbedo, J. G. A.: Factors influencing the use of deep
learning for plant disease recognition. Biosyst. Eng., 172
(2018) 84–91
[8] Brahimi, M., Boukhalfa, K., and Moussaoui, A.: Deep
Learning for Tomato Diseases: Classification and
Symptoms Visualization. Appl. Artif. Intell., 31 (2017)
299–315
[9] Shrivastava, V. K., Pradhan, M. K., Minz, S., and
Thakur, M. P.: Rice plant disease classification using
transfer learning of deep convolution neural network. Int.
Arch. Photogramm. Remote Sens. Spat. Inf. Sci. - ISPRS
Arch., 42 (2019) 631–635
[10] Ozguven, M. M. and Adem, K.: Automatic detection and
classification of leaf spot disease in sugar beet using deep
learning algorithms. Phys. A Stat. Mech. its Appl., 535
(2019) 122537

All rights reserved by www.ijsrd.com 20