AJCS 18(07):408-415 (2024) ISSN:1835-2707

https://doi.org/10.21475/ajcs.24.18.07.p4037

Convolutional neural network as an efficient alternative to supervised

learners for modelling plant diseases

Sridevy Sridarane

, Praveena Sivalogeswaran

, Nirmala Devi

8

1

Department of Physical Science & Information Technology, Tamil

Nadu Agricultural University, Coimbatore-3, India

2

Teaching Faculty, Food and Agribusiness Management, NIFTEM-T,

Thanjavur- 5, India

3

Sri Ramakrishna College of Arts and Science, Coimbatore, Tamil

Coimbatore-3, India

6

Kanchi Mamunivar Government Institute for Postgraduate Studies and

Research, Puducherry -8, India

7

Teaching Assistant, AC & RI, Tamil Nadu Agricultural University,

Coimbatore-3, India

Abstract: The high productivity of the agricultural sector is crucial to

development, amount, and quality of products. Therefore, it is crucial

to find the diseases early in the plant's development. Image

processing is used to identify plant diseases and pests. Artificial

intelligence, machine learning, and visual analysis have all been used

in the past few decades to find and diagnose plant diseases. These

computerized techniques are excellent for quickly monitoring vast

acreage. In this study, we used the foliage of potato plants to identify

diseased leaves affected by early blight, as well as late blight. A

Comparison is thus laid between the Convolutional Neural Network

(CNN) Model for this purpose, and the standard Supervised Learning

training–testing splits are considered for such, and thereby, the

convolutional Neural Network Model is found to be better in

performance than the Supervised Machine Learners irrespective of the

Splitting Ratio of the Training, and the Testing Data.

Keywords: Machine Learning, Supervised Learning, Artificial Neurons,

Solanum Tuberosum.

Introduction

India is a country of farmers; agribusiness and productive employment

account for over sixty per cent of the country's workforce. India's

agricultural industry is reliant on the monsoon months Agriculture

India employs conventional Zhang et al. (2022) techniques and

conventional machinery. This is a result of the destitution of

individuals and ignorance. Farmers planted a variety of crops on their

farms. Here, crops occupy 75% of the land used for cultivation and

commercial crops occupy only the remaining 25%. India has two

cropping seasons – Rabi, and Kharif, and the monsoon is solely

responsible for such. All region's harvest trends are influenced by a

variety of variables, including the environment, climate, the system of

irrigation, financial drivers, facilities, and social engagement.

Agricultural issues are numerous. In India, farming practices remained

resistance to the chemicals, rendering them ineffectual over time. 10%

of the freshwater we use is for cultivation. New industrial methods

need additional water. The manufacturing process of the main crops

in India temporarily stops. Due to an enormous discrepancy amid the

demands and supplies of the nation's growing population and

manufacturing, legislation administrators and administrators are

concerned. On one farming operation, the landowner continually

harvested the same crop, making it less fertile. Studying the

development of vegetation necessitates monitoring of its chemical and

physical characteristics. Chlorophyll Najm et al. (2012) as well as

(2023a). There are two ways we can obtain this knowledge, the

traditional approach, in which producers used their unaided vision to

examine the development of the plants, nevertheless lacked

competence in this field and it is not appropriate for a big region.

Potato Leaves.

farming fields are negatively impacted by diseases that affect plant

leaves. illnesses caused by a variety of microbes, inheritable disorders,

and contagious factors such as bacteria, fungi, and viruses. Huang et

al. (2019) Potato plant leaves have been utilized in this study. The

majority of tuber leaf illnesses are caused by fungi and bacteria. Early

and late forms of blight are bacterium-fungal infections and illnesses.

The most important agricultural product on the planet is the potato. In

India, potatoes are a crop that flourishes in a subtropical environment.

It is economical nourishment since it supplies less expensive celery

for human consumption. Starch, Vitamin B1, along Vitamin C is all

Supervised Learners. Each of these will have its efficiency and

performance to the research problem targeted in this article, but the

most important one would be the algorithm that can predict the

presence of disease-causing pathogens on the surface of tuber with a

higher accuracy than other, and that too without compromising much

on the Spatiotemporal domains. Though similar conduct has been

made in specific samples of potato leaves, in some existing literature,

like Dutta et al. (2023c), and Dutta et al. (2023d) none provokes the

variation in efficacy of the detection of disease in the leaves with

variation in Training Testing Split. Training Testing split is another

and the remaining to compute the efficiency of the model when tested

on data that is completely naïve to them. The sections that follow

declare a brief idea of the nutritional benefits of the Potato tubers,

and their biological domains, demonstrate the methods and data used

for the research, and further delineate the results obtained as a metric

for the comparisons.

Subject of Conduct

-

Solanum tuberosum

Following the grains of rice, wheat, and cereal grains, potatoes

(

Solanum tuberosum

) are the world's fourth-most significant

for potato developers is the creation of cultivars with agronomically

significant features and high preservation

Table 1. Tabularization of the results (Training and Testing F1 Score)

for each of the algorithms used in the subjection of the research split

at a ratio of 80: 20.

for each of the algorithms used in the subjection of the research split

at a ratio of 90: 10.

typically used to evaluate the variation in the genetics of the potato

crop populations, like the analysis of morphology, biochemical

assessment, as well as genomic indicator study. Goodwin et al. (1994)

found limited genetic diversity in processed potatoes; therefore,

molecular indicators are essential for evaluating genetic

diversification. Adequate genetic indicators should be exhibited in all

tissues, organs, and phases of crop growth. They must be at the

chromosomal level. When contrasted with conventional breeding

initiatives, molecular indicators can boost the efficacy as well as the

effectiveness of breeding initiatives. Numerous molecular methods,

The primary constraining aspect of the cultivation of potatoes is heat.

Tuber growth is severely hampered by conditions beneath 10°C (50°F)

in addition to 30°C (86°F). An estimated 320 million tons of potatoes

are produced worldwide annually as per Haas et al. (2009) over an

area of cultivation of around 20 million hectares. Almost thirty per

cent of all tubers originate in both India and China, which are

currently the world's top two potato producers. As opposed to this,

the major potato-producing nations on the continent are Egypt,

Algeria, as well as Morocco, according to the sequence. Since the

beginning of the twentieth century, the cultivation of potatoes has

tubers were produced and consumed in North America, Europe, and

nations in the days when the Soviet Union existed. One of the primary

biological limits on the cultivation of potatoes, especially throughout

tropical climates in addition to some warmer temperate parts of

around the globe, is the presence of pathogenic

bacterial infections. Around seven bacterial infections severely harm

potatoes, particularly their tubers, which are the plant's most precious

resource commercially. The most serious illnesses are thought to be

bacterial wilt and back leg, while minor ones as mentioned by

abnormalities typically appear on older leaves. Early blight may

substantially decrease yields if left unchecked. Eliminating prolonged

moisture on the foliage and using fungicides are the main strategies

for treating this fungal infection. The harmful potato and tomato

ailment referred to as late blight or just potato blight is brought on by

the oomycete, occasionally referred to as water mould, known as

Phytophthora infestans

.

Decision Tree, Random Forest, and Gaussian Naïve Bayes. Further, we

utilized a Convolutional Neural Network to correspond to the

research.

Support Vector Machines (SVMs)

are one of the efficient

supervised learners that can be applied to both regression and

classification-based problems. Finding an axis of separation that

optimally isolates the different categories in the simulation data is the

basic goal of Support Vectorized Machines. The framework for

achieving this is to find a suitable hyperplane with the biggest

collateral, indicating the distinction that exists between the

hyperplane along with the closest data suites for each class. After

Fig 6. Comparative Plot showing the Training, and Testing F1 Scores

for each of the algorithms following 60 – 40 Split.

Fig 7. Comparative Plot showing the Training, and Testing F1 Scores

for each of the algorithms following 50 – 50 Split.

hyperplane that is efficient enough, fresh, new training data can be

categorized by identifying the component of the hyperplane it appears

on. When the data includes a lot of features or whenever there is a

of intrusions, data exploration, as done by Dutta and Saha (2022), and

the recognition of patterns. Due to its non-parameterized nature,

which means that it makes no fundamental presumptions concerning

the pattern of distribution of information, it is frequently used in real-

world circumstances. The artificially intelligent meta-algorithm known

as

, or

Adaptive Boosting

, was created by Yoav Freund and

Robert Schapire, who additionally earned the 2003 Gödel Prize for

their contributions to machine learning. The effectiveness of the

algorithm can be enhanced by combining it with a variety of additional

instructional techniques. The results of the additional algorithmic

learning techniques in particular situations, as raised by Dutta et al.

(2023a). It has been demonstrated that the result of the model

converges to an effective learner even if the learning outcome of every

learner individually is just marginally better compared to arbitrary

speculation. AdaBoost using decision trees to serve as poor learners is

commonly referred to as being the most effective

unconventional classifier. However, every machine learning technique

tends to adapt to specific issue forms more effectively than others,

and it usually includes numerous distinct settings and arrangements

to modify beforehand so it accomplishes its optimum efficiency on an

procedure to ensure that subsequent trees prefer to concentrate on

cases that are more difficult to categorize, as done by Dutta et al.

(2022a). Arguably among the most effective techniques for supervised

learning for both regression as well as classification applications is the

Decision Tree

organizational diagram wherein every node within the structure

symbolizes an evaluation of a characteristic, every branch is a test

result, and every leaf node is a class identifier. A limiting requirement,

such as the deepest possible level of the structure or the least number

of specimens needed for splitting a particular node, is reached by

referred to as estimators, each of whose outputs produce a unique set

of estimates because it is a consolidation technique. The estimators'

estimates are combined by the method of random forest modelling to

yield a more accurate forecast, as referenced by Dutta et al. (2022b).

Gaussian Naive Bayes

is utilized for a variety of classification tasks. The further development

of Naive Bayes is called Gaussian Naive Bayes. It constitutes a single of

Naive Bayes' models for categorization. There is a Gaussian spectrum

for information that is continuous. Because of this, we categorize

continuous data using the Gaussian Naive Bayes. The Gaussian

employed mostly to perform computation and identification of

images. A layer of input, layers that are hidden, and an output tier

make up this structure. The concealed layers of a network of

convolutional neural networks contain any number of layers of

convolution. This typically contains a layer that does a dot product

operation of the input matrices of the layer that corresponds with the

convolutional kernel (refer to Figure 2). Artificial neurons are arranged

in numerous layers to form convolutional neural networks. Synthetic

neurons are computational operations that determine a stimulation

value by summing the weights of various inputs.

in India, the preliminary source of income for a wide group of

individuals is in some way or the other related to the Agricultural

Sector. If we can supportively withstand all of the threats that our

agricultural sector may face, our economy will be directly helped. The

F1 Scores shown in the tables above show the variation of the Training

and the Testing Scores with the variation in the Training and Testing

Split for each of the Algorithms. It was observed that the

Convolutional Neural Networking Model is performing the best for

each of the splitting ratios. Though CNNs have been known to work

well for these cases of plant pathology, the number of training

to get the results reunited in a quite minimal threshold requirement.

This would help us Economically, Ecologically, Socially, Mentally, and

Dutta A, Kumar PK, De A, Kumar P, Harshith J, Soni Y (2023)

Dutta A, Saha M (2022) Contrasting Parallelized with Sequential

Gharib HA, Mandour AM (2022) Effect of Populus nigra spring and

Haas BJ et al. (2009) Genome sequence and analysis of the Irish potato

Huang B, Ruess H, Liang Q, Colleoni C, Spooner DM (2019) Analyses of

Jung MY, Choi DS, Ju JW (2003) A Novel Technique for Limitation of

Kumar PK, Kumar I, Kumar S, Kumar P, Harshith J, Dutta A (2023) Title

Kumar PK, Munjal D, Rani S, Dutta A, Voumik LC, Ramamoorthy A

Smith MR, Martinez T (2011) Improving classification accuracy by

Wang Y, Zheng Y, Liu Y (2022) Identifying vital nodes for influence