Plant disease detection using Deep Convolutional

Neural Network for Corn, Rice and Banana Crop

Divyam Dholwani1, Kaustubh Patil1, Vedangi Thokal1, Pradnya Kulkarni1,

Mamta Bhamare1

1
School of Computer Engineering, Dr. Vishwanath Karad MIT World Peace
University, Pune, Maharashtra

[email protected], [email protected], [email protected],

[email protected], [email protected]

Abstract. Among the most widely farmed and eaten crops globally are
corn, rice, and bananas. However, a variety of both biological and
environmental variables, including weather, pests, soil, and water quality,
can influence these crops, leading to illnesses that have a substantial impact
on crop quality and output. These illnesses not only cause financial loss but
also represent a serious danger to the farming sector. Traditional methods of
detecting diseases are insufficient and frequently depend on laborious eye
examinations.Consequently, having efficient techniques for identifying and
treating these illnesses is crucial. Using a Deep Convolutional Neural
Network for the crops of corn, rice, and bananas, this article tackles this
problem. a layout in which max pooling layers are arranged in a deliberate
manner after convolutional layers with progressively larger filters to
guarantee the identification of certain illness patterns. The findings showed
that the accuracy was 99.7% for Bananas, 98.8% for Rice, and 99.8% for
Corn. This study aims to empower farmers worldwide and support them in
protecting the health and productivity of their crops by showcasing the
application of convolutional neural networks in crop disease identification.
Keywords: Machine Learning, Deep Learning, Neural Networks,
Convolutional Neural Network

1 Introduction

Some of the major obstacles to worldwide agricultural production and food

security include pests and plant diseases Reducing crop losses and promoting
sustainable food production need timely identification and resolution of these
issues.Nevertheless, conventional disease detection approaches, reliant on visual
assessments by human experts, frequently entail labor-intensive processes,
consume significant time, and are susceptible to errors.Over the past few years,
2

the application of machine learning (ML) and deep learning (DL) approaches has
shown remarkable promise in transforming the field of crop illness
detection.These techniques leverage the power of computational algorithms to
analyze and interpret large amounts of data, particularly images of diseased plants,
enabling the accurate and early identification of various plant diseases. Khatoon et
al. [1] provides a comprehensive overview of the latest advancements in deep
learning for plant disease detection. It covers various convolutional neural
network architectures, data enhancement techniques, and disease classification
strategies. Plant health and illness may be recognized with high accuracy using
ML and DL models that are trained to collect and analyze information from
photos, such as its hue, texture, and shape. Saleem, Potgieter and Arif’s extensive
review paper [2], provides a thorough analysis of state-of-the-art deep learning
techniques for plant disease detection and classification.

Convolutional Neural Networks (CNNs) are a unique class of neural networks that
process data with a grid topology, making them ideal for image processing. CNNs
have shown great potential in revolutionizing the field of plant disease detection,
enabling the accurate and prompt recognition of various crop illnesses. By
extracting and analyzing features such as color, texture, and shape from images, In
a recent study conducted by Andrew J et al. [3], the potential of deep learning
techniques, specifically convolutional neural networks (CNNs), was investigated
in terms of their ability to accurately distinguish between healthy and diseased
plants. The study found that CNNs demonstrated great precision in this task. The
researchers managed to accomplish a remarkable 96% accuracy rate,
demonstrating the effectiveness of AI in early disease detection.While Afifi,
Alhumamand and Abdelwahab investigate the use of CNNs for recognizing
various diseases on grape leaves [4]. The developed model achieved a remarkable
99.8% accuracy, highlighting the power of AI in differentiating between healthy
and diseased plant tissues. Simhadri and Kondaveeti [5] focused on developing a
deep learning model for rice disease detection that performs well even in resource-
limited environments with limited computing power and data availability. On the
other hand, we used verified datasets related to these crops to train our model. In
this research, we propose a mechanism that uses CNNs to detect diseases in two
prominent crops, corn and rice. It predicts diseases related to these crops by
processing various images and weather datasets. Furthermore, our work presents
an architecture that combines photos of plants with weather data.. This
architecture utilizes simple pooling operations and convolutions, and is referred to
as CNN augmented with cellular automata rules. This approach aims to effectively
identify plant diseases by making use of the combined information from weather
data and plant images.

This paper is organized into six sections, with section 1 providing the introduction
to the problem, section 2 providing an extensive literature survey, section 3 giving
the design and methodology used in our work, section 4 providing the results and
comparisons to the existing work, section 5 providing the conclusion to our work,
 3

and section 6 discussing the future scope of our research. Seeking an in-depth
understanding of advanced methods and technologies in the field, the goal is to
contribute to the development of more trustworthy and efficient ways for
identifying and controlling pests and plant diseases.

2 Literature Review

In the pursuit of global food security, Kutyauripo et al. [6] conducted a

comprehensive systematic review, evaluating the application of artificial
intelligence (AI) across the entire food production ecosystem. Their analysis of
110 articles highlights AI's diverse role, from crop and cattle production and food
waste control. The research emphasizes AI's potential in enhancing productivity,
ensuring sustainability, and addressing the challenges posed by climate change in
the ever-evolving food system.Addressing the threat of plant diseases impacting
crop yield and economic losses, Muthukannan et al [7] propose a solution
leveraging smart technology. They introduce Feed Forward Neural Network
(FFNN), Learning Vector Quantization (LVQ), and Radial Basis Function
Networks (RBF) to effectively identify and classify diseased plant leaves, with a
focus on beans and bitter gourds. The study demonstrates that FFNN outperforms
other methods, evaluated through simple measures like Accuracy, Precision,
Recall ratio, and F_measure, aiming to enhance crop quality in the Indian
economy. Pokkuluri and Nedunuri developed a novel approach that integrates
Cellular Automata encoding with Convolutional Neural Networks (CNN). The
resulting classifiers, CNN-CA-I and CNN-CA-W, demonstrated remarkable
accuracies of 92.6% and 90.1%, respectively, marking a significant advancement
in the prediction of crop diseases across various crops (Pokkuluri & Nedunuri,
2022).

Reshmi A.M and Prameeja Prasidhan's research [8] stands out for its pioneering
use of an efficient Convolutional Neural Network (CNN) algorithm. Achieving an
impressive 99.5% accuracy, the study employs several methods for image
processing to detect and classify crop diseases, providing valuable insights for
farmers to mitigate losses and enhance productivity.Examining the landscape of
automatic crop disease identification, the study by Justine Boulent et al. (2019) [9]
utilizes Convolutional Neural Networks (CNNs) to analyze 19 relevant works.
With a focus on tomatoes and a predominant leaf-centric approach, the research
highlights the potential of deep learning techniques in advancing practical tools
for sustainable agriculture and food production securityIn their 2023 research,
4

Pawan Shankar Ghodekar et al. [10] present an innovative solution to India's

agricultural challenges. The proposed system, incorporates algorithms for data
preprocessing, CNN model training, and disease classification. Notably, the
project's integration of OpenCV for feature extraction and CNN for predictive
accuracy marks a significant advancement, offering farmers informed decisions on
effective disease management and contributing to the sustainability and
productivity of agriculture. Abade et al. 2020 [11] reviewed 121 papers published
between 2010 and 2019 and found that PlantVillage was the most widely used
dataset, while TensorFlow was the most frequently used framework. Bangari et al.
[12] conducted a review on disease detection using Convolutional Neural
Networks (CNNs), focusing on potato leaf diseases. After examining several
papers, they concluded that CNNs are fairly adept in detection of plant diseases.
Additionally, they discovered that CNNs made a major contribution to a high level
of disease identification. Put more simply, their research shows that CNNs are
highly accurate in spotting potato leaf diseases and are very good at distinguishing
them.

In their comprehensive survey,Tugrul,Elfatimi and Eryigit 2022 [13] , review the

recent surge in applying deep learning, particularly convolutional neural networks
(CNNs), to address plant disease detection in agriculture. Analyzing 100 pertinent
research articles over the last five years, the study emphasizes the efficacy of Deep
Convolutional Neural Networks (DCNN) for prompt disease detection.The
research dives into a number of topics, such as CNN architectures, frameworks,
datasets, and findings from the experiments, and it clarifies why conventional
methods are still widely used.The research not only identifies challenges related to
image background effects but also outlines future directions for advancing plant
disease detection methodologies.In the study conducted by Shelar et al. [14], a
remarkable achievement of 95.6% accuracy was attained using early stopping
during the model training phase, spanning 50 epochs.The research successfully
demonstrates the model's efficacy in detecting and recognizing both healthy and
infected strawberry and potato plants, showcasing the potential for accurate plant
disease identification.Hassan et al. 2021 [15] showcase the potential of deep
convolutional-neural-network (CNN) models for precise crop disease
identification. They innovate by introducing depthwise separable convolution,
cutting down parameters and computational expenses. Trained on a diverse dataset
spanning 14 plant species and 38 disease classes, their models, notably
EfficientNetB0, excel with accuracy rates reaching 99.56%. The streamlined
MobileNetV2 architecture, optimized for mobile devices surpasses traditional
techniques in terms of training duration and accuracy. The study marks a
substantial leap forward, offering promising applications for real-time agricultural
disease detection systems.Srivastava et al. 2020 [16] present a novel approach to
combat plant diseases, employing deep convolutional neural networks (CNNs) for
accurate disease detection in crops. The system streamlines disease recognition
through image classification, replacing time-consuming manual methods. By
 5

gathering and analyzing a diverse leaf image dataset, the authors train and fine-
tune the CNN, achieving an impressive 88% accuracy in overall disease
classification. The proposed methodology, emphasizing simplicity and efficiency,
marks a significant leap in automating plant disease identification, offering
promise for improved precision in agriculture.

3 Data Engineering

Any machine learning model's ability to perform well is closely correlated with
the caliber and variety of the dataset used for training. The leaves of corn, rice,
and banana plants are the main subjects of our research. The creation of the model
and its analysis depend heavily on this part. Researchers and practitioners can
access all of the utilized datasets for free because they are all operated under
public domain licenses.

3.1 Corn Leaf Disease Dataset

It consists of a rich variety of carefully annotated corn leaf images. The images are
categorized into 4 classes on the basis of their diseases, common rust, gray leaf
spot, blight and healthy.

Common Rust. A condition resulting from the fungus 'Puccinia sorghi'.Chlorotic

specks on the leaf surface are one of the early signs of common rust. These specks
grow into tiny, round to elongated, tan patches that might appear in groups. Such
spots appear to be scattered around the adaxial and abaxial leaf surfaces.

Fig. 1. Common Rust in corn leaves. [20]

Gray Leaf Spot. It is caused by ‘Cercospora zeae-maydis’.Two to three weeks

before to tasseling, lower leaves show symptoms for the first time. The lesions of
this disease will appear as yellow or tan rectangular spots, but will expand into
6

large cigar-shaped lesions. Leaves appear grayish in color due to the presence of
fungal spores.

Fig. 2. Gray Leaf Spot. [20]

Blight. It is a disease caused by Exserohilum turcicumis that damages corn. It's

also known as Turcicum leaf blight (TLB). It can cause significant yield loss in
susceptible corn hybrids.Usually, the lowest leaves are the first to show
symptoms.The leaf spots are initially greenish gray and later on turn to pale
brown.They are usually long (upto 15cms). The underside of the leaf under the
lesions produces spores, which give the appearance of dusty green fuzz.

Fig. 3. Corn Blight. [20]

Healthy. Healthy corn leaves are generally green, but can also be purple or
reddish-brown. They are a monocot plant, so they have parallel venation. They
can be 0.3–1.5 meters long and 5–15 centimeters wide.The margins of the leaves
are entire, meaning they don't have obvious teeth. They can also have narrow
longitudinal wrinkles and are sometimes vertically wavy.
 7

Fig. 4. Healthy corn leaves. [20]

3.2 Rice Leaf Diseases Dataset

This dataset consists of a diverse array of diseased rice leaf images.A total of 120
images is present in this dataset. The images are further classified into 3 classes
namely, Leaf smut, brown spot, bacterial leaf blight.

Leaf Smut. Entyloma oryzae is the casuing agent of leaf smut. It's a common but
mild illness that generates little black spots on rice leaves that are angular and
slightly elevated. The spots are linear, rectangular, angular, or elliptical, and
measure 0.5–2.0. On a single leaf, there may be many spots, but they all stay
unique from one another. When the epidermis becomes moist/wet, it breaks open
and releases the black spores. The edges of heavily diseased leaves die and
become pale, while the leaves themselves turn yellow.

Fig. 5. Rice leaf smut. [21]

Brown Spot. It is caused by Cochliobolus miyabeanus.It's also known as

Helminthosporium leaf spot, sesame leaf spot, Helminthosporiose, or fungal
blight. The appearance of several big spots on the leaves—which may eventually
cause the leaf to die entirely—is the most obvious harm caused by this. In cases
where the infection affects the seeds, it results in the formation of unfilled grains
or seeds that are spotted or discolored. Infected seedlings exhibit small, circular
8

lesions that are yellow-brown or brown in color. These lesions have the potential
to encircle the coleoptile (the protective sheath covering the emerging shoot) and
cause distortion in both primary and secondary leaves.

Fig. 6. Rice Brown Spot Leaf. [21]

Bacterial Leaf Blight. It is a highly destructive bacterial disease. It has a

significant impact on the rice cultivation worldwide. It affects millions of hectares
of rice crops each year, the bacterial agent causing this disease is Xanthomonas
oryzae.
Some of the common signs and symptoms are light green to grayish green plants
with water-soaked streaks on the leaves. These streaks form larger yellowish
lesions with uneven edges.Infected leaves change color to grayish green and curl.
As the disease advances, the leaves become yellow or straw-colored and wither.

Fig. 7. Bacterial leaf blight in rice leaf. [21]

3.3 Banana Leaf Diseases Dataset

Numerous illnesses commonly pose a threat to banana agriculture and negatively

affect productivity. These illnesses have a negative impact on banana plants,
stunting their development, reducing their ability to produce fruit, and sometimes
even killing the plants. The dataset includes a wide range of photos that highlight
three common banana leaf spot illnesses called cordana, sigatoka,
pestalotiopsis.For a more thorough inspection, pictures of healthy banana leaves
 9

are also included. So in total 4 classes are taken.

Cordana. It is a fungal disease called cordana leaf spot (CLS) mostly affects
banana trees. Banana diamond leaf spot is the most popular term used to describe
this illness.Two fungi species cause CLS, C. johnstonii and C. musae.. Even
though it is common worlwide, has generally little impact on production.

Fig. 8. Cordana leaf spot on a banana leaf. [22]

Sigotaka. It is a fungal disease that infects the leaves, causing spots that can lead
to leaf drying and defoliation. This can affect the quality of the fruit and reduce
yield. The fungus Cercospora musae is the cause of sigatoka, often referred to as
Cercospora leaf spot. It is one of the most harmful banana illnesses common in
banana plants across the world.

Fig. 9. Banana leaf affected by sigotaka. [22]

Pestalotiopsis. Pestalotiopsis palmarum is a fungus that causes a disease in

bananas, coconuts, and date palms. The fungus is responsible for the formation of
spots on the leaves, blight on the petiole/rachis, and occasionally, bud rot in
palms. Pestalotiopsis palmarum can attack every part of the leaf, starting from the
10

base and extending to the tip. Pestalotiopsis disseminata is a species of fungi in

the family Pestalotiopsidaceae that causes banana leaf spot.

Fig. 10. Pestalotiopsis in banana leaves. [22]

Healthy. Healthy banana leaves are large, wide, and slightly rounded with a
smooth surface. They are waxy, flexible, and glossy, and can grow up to two
meters long and half a meter wide. Healthy leaves are generally dark green in
color, flexible and intact with smooth edges.

Fig. 11. Healthy banana leaves. [22]

Image characteristics. As a part of pre-processing, images were re-sized,

OpenCV library in Python was used to resize our images and retrieve their shape.
The desired dimensions for our output images in this case were 180px in height
and width.

Data Splitting. Two divisions were made in the dataset, Training set and the
Testing set. The division for the disease detection model was 75-25,implying that
75% of the dataset was used for training, while the remaining 25% was set aside
for testing the model. The rationale behind this approach was to ensure sufficient
data for training while also having a substantial set for unbiased evaluation. The
 11

split was executed by randomly shuffling the dataset and subsequently partitioning
it into two disjoint sets. This methodology allowed for a balanced consideration of
training effectiveness and the model's capacity to generalize unseen data.

Table 1. Summary of our datasets’ characteristics

Dataset Classes Number of Total Images
images per class
Corn Leaf
Common Rust, Gray Leaf Spot, 1306, 574, 1146,
Disease 4188
Blight, Healthy 1162
Dataset

Rice Leaf
Leaf smut, Brown spot, Bacterial
Diseases 40 each 120
leaf blight
Dataset

Banana Leaf Sigotaka,Cordana,Pestalotiopsis, 400 each 1600

Spot Diseases Healthy
Dataset

4 Proposed Model and Methodology

The methodology for Disease Detection in plants involves leveraging Deep

Learning models to successfully identify diseases in plants based on input
images.Neural Networks are an essential aspect of Deep Learning models. They
consist of several layers of nodes - one convolutional “input layer”, other hidden
layers, also termed max pooling layers and an output layer. Every nodeis
connected to another. All nodes have an associated weight and threshold, forming
a Neural Net. Among the various types of Neural Nets exist Convolutional Neural
Networks (CNN). Main applications of CNNs inlcude image and video detection.
Prior to CNNs, object detection relied on manual and extensive feature extraction
techniques. Nevertheless, Convolutional Neural Networks (CNNs) implement
concepts from linear algebra, specifically involving matrix convolution and
multiplication., to detect patterns within images, making it more adaptable. CNNs
stand out from other Neural Networks due to their excellent performance with
audiovisual inputs. Three different types of layers make up CNNs: the
Convolutional Layer, Pooling Layer, and Fully-connected Layer (also known as
the Dense Layer). The network is started with the convolutional layer, which may
be followed by other convolutional or pooling layers. The focus of the first layers
is on fundamental elements such as colors and edges. As the layers proceed, the
12

CNN becomes more complex, eventually dealing with more significant elements
of the object till it successfully performs identification. [17]

Fig. 12. Convolutional Neural Network Architechture

To understand the effective use of CNNs in agriculture applications, a systematic

literature review was conducted. The review aimed to gather insights from peer-
reviewed research papers available on various platforms. Articles published before
the year 2000 were filtered out in order to maintain relevancy.Preparing a
Convolution Neural Network organization for pattern detection using the dataset
was proposed. The model was tailored sequentially, implying a linear stack of
layers. Initially, three convolutional layers are integrated. They apply
convolutional operations, aiming to extract features from the images. They all
have identical padding and are set up with a 3x3 filter size.to preserve the input
dimensions. The number of filters increases progressively, to enhance detection of
crop-specific disease patterns. Max-pooling layers come after every convolutional
layer. It downsamples the spatial dimensions while preserving the most important
data for further processing. As a result, the computational load is decreased, and
the subsequent layers are more focused on essential features, hence streamlining
the learning process. This downsampling technique contributed to our model
efficiency by minimizing redundant details and noise. It selects the maximum
values within a localized group of pixels hence retaining salient features. Strategic
integration of pooling layers between convolutional layers is a key element of the
model. Subsequently, a flatten layer is utilized. This layer flattens the
multidimensional output obtained from its predecessor layers into a one-
dimensional array, so that the data is prepared to be input into densely connected
layers. Finally, two fully connected or dense layers are utilized. The first layer
comprises 128 neurons activated by the rectified linear unit activation function
and serves as intermediate representation. ReLU is one of the most commonly
used activation functions. It removes negative values from the training model.
 13

[18]. The final layer aligns with the number of classes corresponding to the
specific crop under consideration. The model is purposefully tailored in a
hierarchical fashion. The data then moves through the layers causing the model to
gradually evolve its understanding from basic features such as colours, edges to
more advanced elements providing a comprehensive analysis of plant health. Its
progressive nature captures the patterns that are relevant to the health of crops and
offers a thorough method for detecting infections in bananas, rice, and maize early
on. This design helped us capture the nuanced patterns specific to crop diseases.

Fig. 13. CNN Model Representation

5 Results and Discussion

This research work proposes the use of Convolutional Neural Network for disease
detection in corn, rice and banana plants. The models will be trained on a dataset
obtained from the Ministry of Statistics and Programme Implementation of India .
A model has to be tested using fictitious data in order to be evaluated for efficacy.
All of the input values in this testing data are utilized to assess the model. The
model's performance is further elucidated by comparing its output predictions with
the expected results. , the testing data encompasses the most recent year, while the
remaining data is used for training purposes.

There are 2 graphs shown for each crop, ‘Model Accuracy’ and ‘Model Loss’.
14

The accuracy of the model was calculated by using the following formula.

Number of correct predictions

Accuracy =
Total number of predictions
(1)

and the loss was calculated using the formula.

n
LCE = − ∑ t i log ( p i ), for n classes, (2)
i=0

where ti is the truth table and pi is the Softmax probablity for the ith class [23].

The left graph shows the model's accuracy over time, measured as the fraction of
crops the model correctly identified as diseased or healthy. The model's loss over
time, which is a gauge of how inaccurate the model's predictions are, is displayed
on the right graph.The quantity of training epochs is indicated on the x-axis.An
epoch denotes the end of an iteration over the training set. The model is exposed
to all of the training examples during each epoch, and it modifies its parameters in
response to the patterns it discovers and gains knowledge from the data.

5.1 Rice

Fig. 14. Model accuracy and loss during training for rice disease detection

The accuracy graph starts around 0.38, calculated using equation 1, indicating that
the model correctly identified about 38% of the rice plants as diseased at the
beginning.The accuracy then continues to rise steadily and finally peaking at 30
epochs.The loss in data starts around 3.0, calculated using equation 2.Over the
course of the 30 epochs, the loss line drops dramatically, reaching a value of 0.007
at epoch 30. This indicates that as the model learnt, its prediction errors decreased
dramatically.
 15

5.2 Banana

Fig. 15. Model accuracy and loss during training for banana disease detection

The accuracy graph starts around 0.5, calculated using equation 1, indicating that
the model correctly identified about 70% of the banana plants as diseased at the
beginning.The accuracy line rises steadily over the 30 epochs, reaching a peak of
0.99 (99%) at epoch 30. The loss graph starts around 1.2, calculated using
equation 2, shows how far off the model's initial estimates were from the right
answers.The loss line then falls sharply over the 30 epochs, reaching a value close
to 0.007 at epoch 30.

5.3 Corn
16

Fig. 16. Model accuracy and loss during training for corn disease detection

The accuracy graph begins at 0.24, calculated using equation 1, meaning that
around 24% of the rice plants were initially accurately recognized by the model as
unhealthy.After then, the accuracy increases gradually before peaking at 30
epochs.About 0.7 is where the data loss begins.The loss line experiences a
significant decline over the course of the 30 epochs, peaking at 0.005 at epoch 30,
calulated using equation 2.

Thus, the results observed are a testimony for the potential application of this
CNN model in real-world scenarios to aid in early and accurate diagnosis of plant
diseases.

Table 2. Model’s accuracy and data loss for each crop after 30 epochs
Crop Name Accuracy Data Loss

Rice 0.9889 0.0104

Banana 0.9975 0.0052

Corn 0.9986 0.0075

5 Conclusion

The proposed research suggests a convolutional neural network approach for

disease detection in corn, rice and banana. Trained on a dataset obtained from the
Ministry of Statistics and Programme Implementation of India, the model exhibits
significant accuracy in identifying diseases within these crops. The foundation of
neural networks, explained through nodes layers and provide a fundamental
 17

structure to our solution. The CNN was tailored to include convolutional layers
with increasing filter size, followed by strategic placement of max pooling layers
that minimize the spatial dimensions of the input. Due to this the learning curve of
the model was improved drastically. The hierarchical design resulted in highly
dense layers, which proved to be instrumental in capturing nuanced patterns
specific to crop diseases. Through meticulous testing and evaluation, the model
displayed remarkable efficiency. The learning capabilities of the model are
displayed via the accuracy and loss graphs, which indicate the performance peak
at 30 epochs. The model showed staggering accuracy of 98.8% for Rice, 99.7%
for Banana and 99.8% for Corn.

6 Future Scope

Looking ahead, the future implications of this research are expansive. This
includes scaling the model to encompass a wide range of crops will ensure the
versatility of the system. The development of real time monitoring systems with
user-friendly interfaces holds the power to empower farmers, allowing them to see
instant and actionable insights. Collaborative efforts with agricultural institutions
will contribute to the real-world implementation of the model. This trajectory not
only promises further technological advancements in the agriculture sector, but
also a positive impact on global agriculture.

References

1. Shaheen Khatoon1, Md Maruf Hasan1, Amna Asif1, Majed

Alshmari1and Yun-Kiam Yap (2020). Image-Based Automatic
Diagnostic System for Tomato PlantsUsing Deep Learning.
2. Muhammad Hammad Saleem, Johan Potgieter and Khalid Mahmood
Arif (2019) Plant Disease Detection and Classification by Deep
Learning.
3. Andrew J., Jennifer Eunice, Daniela Elena Popescu, M. Kalpana
Chowdary and Jude Hemanth (2022) Deep Learning-Based Leaf Disease
Detection in Crops Using Images for Agricultural Applications
4. Ahmed Afifi, Abdulaziz Alhumam and Amira Abdelwahab (2021).
Convolutional Neural Network for Automatic Identification of Plant
Diseases with Limited Data.
5. Chinna Gopi Simhadri and Hari Kishan Kondaveeti (2023). Automatic
Recognition of Rice Leaf Diseases Using Transfer Learning.
6. Innocent Kutyauripo, Munyaradzi Rushambwa, Lyndah Chiwazi (2022).
Artificial intelligence applications in the agrifood sectors.
18

7. K. Muthukannan, P. Latha, R. Pon Selvi and P. Nisha (2015)

Classification Of Diseased Plant Leaves Using Neural Network
Algorithms. vol. 10, no. 4, march 2015 ISSN 1819-6608
8. Reshmi A.M, Prameeja Prasidhan (2022). Leaf Disease Detection Using
CNN. IJCRT Volume 10, Issue 3 March 2022 ISSN: 2320-2882
9. Justine Boulent, Samuel Fouche,r Jérôme Théau, Pierre-Luc St-Charles
(2019) Convolutional Neural Networks for the Automatic Identification
of Plant Diseases. Front. Plant Sci. Sec. Technical Advances in Plant
Science Volume 10 - 2019
10. Pawan Shankar Ghodekar, Vishwajeet Yermune, Aditya Sable, Ranjit
Mandhare, Madhavi Bhosale (2023) Plant Leaf Disease Detection Using
Cnn. International Research Journal of Modernization in Engineering
Technology and Science Volume:05/Issue:04
11. Andre S. Abade, Paulo Afonso Ferreira, Flavio de Barros Vidal (2020)
Plant Diseases recognition on images using Convolutional Neural
Networks: A Systematic Review.
12. Choyon Chandra Bonik, Flora Akter, Md Harunur Rashid, Abdus Sattar
(2023) A Convolutional Neural Network Based Potato Leaf Diseases
Detection Using Sequential Model. 2023 International Conference for
Advancement in Technology (ICONAT).
13. Bulent Tugrul, Elhoucine Elfatimi and Recep Eryigit (2022)
Convolutional Neural Networks in Detection of Plant Leaf Diseases: A
Review. Advances in Agricultural Engineering Technologies and
Application ISSN 2077-0472
14. Nishant Shelar, Suraj Shinde, Shubham Sawant, Shreyash Dhumal, and
Kausar Fakir (2022) Plant Disease Detection Using Cnn. ITM Web of
Conferences 44, 03049.
15. Sk Mahmudul Hassan, Arnab Kumar Maji, Michał Jasiński, Zbigniew
Leonowicz and Elżbieta Jasińska (2021) Identification of Plant-Leaf
Diseases Using CNN and Transfer-Learning Approach.
16. Prakanshu Srivastava, Kritika Mishra, Vibhav Awasthi, Vivek Kumar
Sahu and Mr. Pawan Kumar Pal (2020) Plant Disease Detection Using
Convolutional Neural Network. International Journal of Advanced
Research (ISJR) ISSN: 2320-5407
17. Reem Ibrahim Hasan, Suhaila Mohd Yusuf and Laith Alzubaidi (2020)
Review of the State of the Art of Deep Learning for Plant Diseases: A
Broad Analysis and Discussion.
18. Guifang Lin, Wei Shen (2018) Research On Convolutional Neural
Network Based On Improved Relu Piecewise Activation Function.
19. International Congress of Information and Communication Technology.
20. Corn leaf disease dataset,
https://www.kaggle.com/datasets/unknown6874/corn-leaf-disease-
datasethttps://www.kaggle.com/datasets/unknown6874/corn-leaf-disease-
dataset, last accessed 2023/12/22
21. Rice leaf disease dataset,
https://www.kaggle.com/datasets/vbookshelf/rice-leaf-diseases, last
 19

accessed 2023/12/22
22. Banana leaf disease dataset,
https://www.kaggle.com/datasets/shifatearman/bananalsd, last accessed
2023/12/22
23. Boudjemaa Boudaa,Hadil Touhami (2023) A Graph ATtention Networks
Model for Session-Based Recommender Systems