Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)

IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

Prediction of Pathogen Causing Rice Plant Disease

and Recommendation using Enhanced Machine
2024 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC) | 979-8-3503-7519-0/24/$31.00 ©2024 IEEE | DOI: 10.1109/ICAAIC60222.2024.10575163

Learning Technique
1st J Suji Priya 2nd M Hema Priya 3rd M Iyswarya
Department of Master of Department of Master of Department of Master of
Computer Applications Computer Applications Computer Applications
Sona College of Technology Sona College of Technology Sona College of Technology
Salem, India Salem, India Salem, India
Sujipriyaj@gmail .com [email protected] m iyswaryamathi0903@g mail.co m

4th K Kiruthika
Department of Master ofComputer Applications
Sona College of Technology
Salem, India
kiruthikakarunakaran18@g mail.co m

Eastern civilizations have a long history of cultivating rice.

Abstract: Plant diseases pose significant challenges to
agricultural productivity, impacting both the quality and In tropical countries, the process of growing rice is carried
quantity of crop yields. Early detection and effective out under a range of meteorological conditions, from the
management of these diseases are essential for mitigating their wettest to the driest. Because of the current state of disease-
detrimental effects on agricultural output. However, related issues, there are many diseases that have a direct
traditional methods of disease monitoring and diagnosis are impact on the nation's rice output. Numerous diseases exist
often costly and require continuous expert intervention. To directly that affect the country's rice output because of the
address these challenges, this study proposes a novel approach existing status of disease- related concerns. Previous
that combines image processing techniques and advanced research has suggested that the main causes of this annual
machine learning methodologies for the accurate identification decrease in rice yield include viruses, fungus, and bacteria-
and classification of rice leaf diseases. By utilizing datasets related illnesses. These diseases of the rice crop tend to
comprising images of diseased rice leaves, environmental data, reduce rice output annually by about 37%. However, a
and historical disease reports, the proposed system utilizes significant number of individuals, similar to those in the
Convolutional Neural Networks (CNNs) for image analysis and past, use rice cultivation as a life strategy; as a result, the bulk
feature extraction. S ubsequently, a S upport Vector Machine of those people are familiar with the practice. Now, though,
(S VM) classifier is employed for disease detection and things are a little different because younger people are less
classification based on the extracted features. The integration aware of these diseases; therefore, the government is giving
of image processing and machine learning technologies offers a agricultural teachers knowledge about them. There are so
cost-effective and efficient solution for early disease detection
many diseases that sometimes even the teachers are unable to
and management in agriculture. By automating the process of
disease identification, the proposed system facilitates timely identify the specific ailment without visiting the fields, thus it
interventions, thereby minimizing crop losses and promoting can be a difficult work. When considering all these things,
agricultural sustainability. these want to know some efficient method to minimize this
issue. To rectify these issues, this project makes use of the
Keywords: Machine Learning, prediction, CNN, Rice Leaf technologies available today to address these concerns. Rice,
Diseases, Autoencoder, Fertilizer Recommendation. a fundamental staple food for billions of people worldwide,
serves as a major component of daily meals and a crucial
I. INTRODUCTION source of energy and nutrition. With a rich history deeply
rooted in Eastern civilizations, rice cultivation has been a
Rice is an important food for many people around the longstanding practice, adapting to a wide range of
world. People eat rice as a major part of their meals, and it's a meteorological conditions in tropical countries.
basic staple food for them. It provides a lot of energy and is a
key source of nutrition for billions of people globally.
However, the current state of disease-related concerns has

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 643

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on February 01,2025 at 07:39:15 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

emerged as a significant threat to the global rice output. In challenge as younger generations are less acquainted with the
recent times, the impact of diseases on rice production has specific diseases affecting crops. Recognizing the need for
become a pressing issue, leading to a considerable annual intervention, governments are taking proactive measures, such
decrease in yield. The primary culprits identified in previous as providing agricultural teachers with knowledge about
research include viruses, fungi, and bacteria-related illnesses, prevalent rice diseases. However, the complexity of
collectively contributing to a substantial 37% reduction in rice identifying these ailments, sometimes requiring field visits
output. Despite the historical significance of rice cultivation as for accurate diagnosis, poses a considerable challenge even for
a life strategy for many communities, there is a growing experienced educators.
image and extract the specified region of interest using an
IOT- enabled camera. ROI extraction enhances the image's
After identification, prediction can be done by a
quality and eliminates extraneous information. It computes an
diverse dataset comprising images of healthy and diseased image's feature set. Finally, the system utilizes SVM, a two-
rice leaves, along with relevant environmental data and
step process that entails training and testing, to carry out the
historical disease records, is collected and pre-processed. By classification. It has an accuracy of 92% [6]. Plant disease
integrating technological solutions, the aim is to enhance
classification and detection dataset gathered from
disease detection, improve awareness among cultivators, and PlantVillage dataset in Kaggle. It includes pictures of 15
ultimately contribute to the sustainability of rice cultivation
distinct plant classes potato, pepper, and tomato and their
in the face of evolving agricultural challenges.
leaves. The dataset was split up into three smaller datasets,
each of which was subjected to ConvNet analysis. For the
II. RELATED WORKS detection of tomato, pepper, and potato plant diseases, we
obtained accuracy rates of 98.3%, 98.5%, and 95%.
According to experimental data, the model was successful in
By Enhanced Machine learning based rice leaf disease detecting and classifying plant leaf diseases with a high
prediction system follows all the necessary pre-processing, a degree of accuracy [7].
variety of machine learning methods, such as KNN (K-
Considering the leaves, check the environmental stress and
Nearest Neighbor), Decision Tree, Naive Bayes, and Logistic
soil fertility by measuring moisture, salinity, etc. Drones
Regression, were used to train the dataset. Using the test
equipped with cameras would monitor the crops, and pictures
dataset, the decision tree algorithm produced an accuracy of
would be taken on a regular basis. The system would operate
more than 95% [1]. The plant disease detection can be done
on datasets of healthy and diseased bottle gourd, maize, and
using deep leaning CNN algorithm. For the purpose of papaya leaves that were collected in a controlled
representing healthy, diseased, and snail-infested plants, a
environment. The severity of the disease can be predicted by
three-class classifier was developed. A total of 600 photos the Convolutional Neural Network [8]. The identification of
were used for the training. The accuracy achieved here
agricultural diseases through the use of deep learning,
91.23% [2]. Brown spot disease in paddy crop can be
machine learning, hyper spectral image analysis, the Internet
detected by K-means clustering method, which has been used
of Things, and image processing techniques. A comparative
to denoise the image into three images bas ed on colour.
analysis of various methods used to identify agricultural
Disease intensity can be determined by applying image diseases was done. This study also addresses the various
denoising [3].
obstacles that need to be discussed and alternative solutions
It is suggested to use the 3-means clustering approach [9].
or K-means clustering to identify the initiation stages on the
Tomato leaf photos are used by the deep learning
leaves of rice crops. The disease can be detected by dead
architecture, and a convolutional neural network is also used to
patches and the yellowing of the leaves. This approach is a
classify tomato diseases. The models' performance for both
very effective way to identify any defective parts on rice crop
healthy and different kinds of unhealthy leaves is also reported.
leaves [4]. Diseases are predicted in the mushroom and It concludes that when trained with deeper networks on
soybean datasets using numerical or categorical variables.
segmented images, all the architectures performed better in
Artificial Neural Network, Naïve Bayes, k-Nearest classifying the disorders. The effectiveness of every
Neighbour, Support Vector Machine, Decision Tree, and
experimental study that is included in this work [10].
Random Forest are some of the methods that can be used for
Agricultural tasks such as forecasting the price of agricultural
classification and prediction. The goal of this work is to
commodities on the market prior to planting, choosing the best
propose classification techniques that may be applied to
cultivars to plant, figuring out the ideal time to plant, assessing
datasets with raw measurements rather than images in order
supply and demand risk, and prioritizing investments.
to classify and predict diseases [5].
Additionally, it helps farmers with tasks like crop management,
It proposes an IoT, ML, and image analysis -based which includes activities for predicting crop quality, yield,
method for recognizing the infection. Take the required disease, weed, and growth, among other things [11].

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 644

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on February 01,2025 at 07:39:15 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

The generated host-pathogen networks are shown in making, this centralized database facilitates the collective
an enhanced display with external links to other protein efforts towards optimizing rice field management practices .
annotation sites. Furthermore, there are numerous new and
established host-pathogen systems that might be studied
further to improve our comprehension of intricate disease-
causing agents. The created models are set upon an openly
accessible web server [12]. With the help of images from the
PlantVillage dataset, which botanists have also identified with
four severity stages as ground truth, a group of convolutional
neural networks is trained to evaluate the severity of the apple
black spot with a 90.4% total accuracy on the hold-out test
set, the deep learning is the best model [13].
The classification of leaf images supports a disease
recognition model. The Convolutional neural networks can be
used for image processing in order to detect plant diseases.
CNN is designed specifically to handle image recognition and
pixel input [14]. By using Shift-Invariant Base Pass Domain
Filter in conjunction with Rapid Bilateral Filtering to maintain
edges and reduce noise sensitivity, the DTID framework Fig 1: Architecture Diagram
improves contrast in segmented pictures. Affine Planar
Transformation is incorporated to optimize the quality of The entire process of applying machine learning to detect leaf
contrast enhancement. DTID improves enhancement quality disease is explained in detail. The entire detection process is
by 27% and reduces filtering time by 54% when compared to broken down into steps, from gathering datasets to identifying
other approaches like GUMA, HMRF, SWT, and EHS. diseases. Plant disease diagnosis is a multi-step process that
Additionally, improving contrast quality by 28%, and raising begins with gathering data from multiple sources. After that,
detection accuracy rate by 26% [15]. preprocess it to correspond with the network's input
requirements. This technology uses transfer learning to
improve accuracy through efficient network architecture. A
III. PROPOSED WORK classifier is used to detect the disease after the features of the
image have been extracted.
The proposed system architecture serves as the foundational Data Preprocessing is an important element of data analysis,
framework explaining the structure, functionalities, and especially for data mining and machine learning. It describes
diverse perspectives of a given system. This holistic the procedure for cleaning, converting, and getting raw data
framework shown in fig. 1, integrates cutting-edge ready for analysis. Before it can be evaluated, the raw data that
technologies such as advanced sensors and Internet of was gathered from multiple sources may include errors,
Things (IoT) devices, aimed at facilitating real-time missing numbers, outliers, and other irregularities. The
monitoring of rice fields. Within this architecture, machine procedures are involved in data preparation to get the data
learning and computer vision algorithms are seamlessly ready for analysis.
integrated to discern and categorize diseases, thereby
furnishing timely insights crucial for effective field Image Denoising is the process of dividing an advanced picture
into different parts is called picture denoising. Grouping
management. SVMs are well-suited for binary classification
tasks and can complement CNNs in disease identification . images of high quality completes the partitioning process.
To enhance accessibility and usability for farmers, a user- Smaller- sized and low-resolution photos were not thought to
centric mobile application or web portal is proposed. This be the best choices for the dataset throughout the image
platform is designed to provide comprehensive information collection process. Therefore, only larger and higher-resolution
on prevalent rice diseases, including their symptoms and photos are taken into account during the entire process.
recommended mitigation strategies. Additionally, the Viruses, fungi, bacteria, and other insects can harm leaves.
application incorporates push notification features to deliver The proposed disease detection module uses a multi-class
timely alerts to users. Collecting a diverse dataset and convolutional neural network approach to identify if a leaf
standardizing preprocessing techniques ensure data quality image is normal or damaged. Depend on the color, structure,
and availability. For this centralized database infrastructure and texture of leaf layers are built. The network's
is implemented. This repository enables the efficient storage convolutional layers extract features like edges, textures, and
and dissemination of disease-related data among various shapes from the input images. The fully connected layers
stakeholders, including researchers, domain experts, and then process these features to make predictions,
farmers. By fostering collaboration and informed decision - distinguishing between healthy and diseased leaves based on

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 645

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on February 01,2025 at 07:39:15 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

learned representations. affected leaves. For example, copper oxychloride is suggested

for leaf smut, mancozeb spray is advised for brown spots, and
-sklearn: A machine learning library providing various tools
steptomysin sulphate is recommended for bacterial leaf blight
for data mining and data analysis. It is used here for creating disease. Organic fertilizers are used. This knowledge is
the SVM model, data splitting, and evaluation.
essential for figuring out the nutritional needs of plants and
-skimage: Image processing library for reading, resizing anticipating any deficiencies that could lead to disease.
images.
Function: load_image_filesThis function aims to load image IV. RESULTS AND DISCUSSION
files from a specified directory, preprocess them, and
encapsulate them for machine learning purposes. The This implementation focuses on utilizing advanced machine
Parameters involved are: learning techniques to predict pathogens causing rice plant
container_path: The file path to the directory containing diseases and provides recommendations for effective disease
image data, organized into subdirectories by category. management. Remarkably, the project achieved a high
dimension: Target dimension for resizing images, defaulting average classification accuracy exceeding 98%, Proper
to 104x104 pixels. validation and evaluation procedures, including cross -
validation, ensure accurate assessment of the model's
The process involved here is to iterate through directories in performance. Continual monitoring and updating with new
the given path, assuming each directory's name represents a data and improvements maintain accuracy over time. The
category (label) for the images it contains. For each image recommendations generated by the machine learning models
found, it reads and resizes the image to the specified closely align with established agricultural practices. By
dimension, using anti-aliasing to minimize distortion. The identifying and classifying various pathogens responsible for
resized image is then flattened (converted from a matrix of leaf illnesses, the model demonstrated its potential to assist
pixel values to a vector) and added to the dataset, alongside its farmers in early disease detection with high accuracy,
target label (derived from the enclosing directory's name).This precision, and recall rates. The work represents a significant
returns a Bunch object (similar to a Python dictionary but step towards more efficient and data- driven approaches to
allows access to keys as attributes), containing: address the complex challenges of rice plant diseases. The
- data: Flattened image data. findings underscore the importance of integrating technology
into traditional farming practices, leading to more sustainable
- target: Numeric labels indicating the class/category of each and productive agricultural systems. Overall, this work
image. contributes valuable insights and practical recommendations
for enhancing the resilience and productivity of rice crops
- target_names: Names of each category.
through innovative, data-driven approaches to disease
- images: The original (resized) images, prior to flattening. management.
load_image(path) seems to be a function call that loads an
image from a provided path. This image will likely be used to
make a prediction. If the model file is found and loaded
successfully, the code executes clf.predict(img) to predict
the disease based on the image data. clf is likely a
classification model. If the model file cannot be found
('Model not Found'), it notifies the user that the model is
missing. After 5000 milliseconds (5 seconds), this
notification label is destroyed by calling
destroy_widget(noti).

After predicting the disease, the names of the diseases is

displayed. The prediction result pred is converted to strings,
concatenated, and then converted to an integer to index the
labels array (a = int("".join(s))).This label containing the Fig2: Home page
prediction information is placed and set to be des troyed after In Fig.2 the home page of the project was created to upload
7000 milliseconds (7 seconds), along with several other the images, which have been stored in the local drivers.
widgets, such as display, imageFrame, dn1.
Depending on the severity of the disease, the fertilizer
recommendation module offers fertilizer suggestions for

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 646

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on February 01,2025 at 07:39:15 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

Fig 3: Uploading Rice Leaf Image Fig 6: Fertilizer Recommendation

The leaf images are collected and stored on the local disk.
Image Fig.5 and Fig.6 shows the predictions of defective
By choosing the file and leaf image, it gets uploaded. The
leaf, it will displays the exact disease of the leaf, as well as
Fig. 3 shows the leaf image gets uploaded.
the fertilizer and amount of fertilizer to be used.

V. CONCLUSION
This work employs image preprocessing of leaves
coupled with Convolutional Neural Networks (CNNs) to
achieve accurate identification and classification of plant
diseases. CNNs, recognized as the most effective method for
image data prediction and enhance the precision of disease
classification. Beyond identification, the research work takes a
significant step forward by recommending the appropriate type
and quantity of fertilizer for the affected plants, thereby
offering a practical solution to improve crop health and yield.
Looking ahead, this implementation envisions more effective
Fig 4: Healthy leaf Image detection of plant diseases affecting larger leaves through the
integration of emerging technologies. To operationalize these
Fig.4 shows the image of healthy leaf. After the process of advancements, the establishment of a real-time data collection
the CNN algorithm, it predicts the healthy leaf, if the leaf is system from farms is recommended, integrating technologies
not affected by any disease, and it displays the message like drones, Internet of Things devices, and smartphone apps.
"HealthyLeaf". This system would empower farmers to submit photos and
data for instant analysis and receive timely suggestions. A
user-friendly smartphone app is proposed to complement this
system, allowing farmers to snap photos of diseased plants and
receive prompt diagnoses along with personalized
recommendations. In essence, this work lays the groundwork
for a comprehensive, technology- driven approach to plant
disease management, promising increased agricultural
productivity and sustainability. By leveraging CNNs' ability to
extract features and patterns from images. Through a process
of image analysis and classification, the system distinguishes
between healthy and diseased leaves with high accuracy and
this algorithm identifies defective parts based on predefined
disease symptoms. This system distinguishes between healthy
and diseased leaves with high accuracy. CNN enhances the
Fig 5: Disease Detection Image accuracy of disease classification, especially across a diverse
range of fruits and vegetables. Ensemble learning, combining
predictions from multiple models, reduces the risk of
overfitting and boosts overall performance.

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 647

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on February 01,2025 at 07:39:15 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC-2024)
IEEE Xplore Part Number: CFP24BC3-ART; ISBN: 979-8-3503-7519-0

REFERENCES

[1] K. Ahmed, T. R. Shahidi, S. M. Irfanul Alam, and S. Momen, “Rice leaf

disease detection using machine learning techniques,” in Proceedings of the
2019 International Conference on. Sustainable T echnologies for Industry
4.0, Dhaka, Bangladesh, December 2019.
[2].R. R. Atole and D. Park, “ A multiclass deep convolutional neural
network classifier for detection of common rice plant anomalies,”
International Journal of Advanced Computer Science and Applications, vol.
9, pp. 67–70, 2022.
[3]. R. Islam and M. Rafiqul, “An image processing technique to calculate
percentage of disease affected pixels of paddy leaf,” International Journal of
Computer Application, vol. 123, pp. 28–34, 2022.
[4]. P. Kumar, B. Negi, and N. Bhoi, “Detection of healthy and defected
diseased leaf of rice Crop using K-means clustering technique,”
International Journal of Computer Application, vol. 157, no. 24–27, 2020.
[5]. Maria Morgan, Carla Blank, Raed Seetan ,” Plant disease prediction
using classification algorithms” Distributed by IAES International Journal of
Artificial Intelligence 10 (1), 257, 2021.

[6] S .Vijayalakshmi , G. Balakrishnan , S. Nithya Lakshmi,” Early

Prediction of Plant Disease Using AI Enabled IOT “Distributed by Data
Science and Security: Proceedings of IDSCS 2021, 303 -309, 2021.

[7] A Lakshmanarao, M Raja Babu, T Srinivasa Ravi Kiran,” Plant disease

prediction and classification using deep learning ConvNets” Distributed by
2021 International Conference on Artificial Intelligence and Machine Vision
(AIMV), 1-6, 2021.

[8] Akash Kumar Gupta, Kishan Gupta, Jayant Jadhav, RugvedV

Deolekar, Amit Nerurkar, Sachin Deshpande “Plant disease prediction
using deep learning and IoT ” Distributed by 2019 6th International
Conference on Computing for Sustainable Global Development
(INDIACom), 902-907, 2019.

[9] Asraa Safaa Ahmed, Zainab Kadhm Obeas, BA Alhade, Refed

Adnan Jaleel “ On using artificial intelligence and the internet of things
for crop disease detection: A contemporary survey” Distributed by
Agriculture 12 (1), 9, 2021.

[10] Muhammad EH Chowdhury, T awsifur Rahman, Amith

Khandakar, Mohamed Arselene Ayari, Aftab Ullah Khan, Muhammad
Salman Khan, Nasser Al-Emadi, Mamun Bin Ibne Reaz, Mohammad
T ariqul Islam, Sawal Hamid Md Ali,” Automatic and reliable leaf
disease detection using deep learning techniques” Distributed by
AgriEngineering 3 (2), 294-312, 2021.

[11] Pooja Akulwar,”A recommended system for crop disease detection

and yield prediction using machine learning approach”, Agricultural and
other industries,141-163,2020.

[12] Rakesh kaundal ,Cristian D Loaiza, Naveen Duhan,Nicholas

Flann,”A comprehensive deep learning platform for accurate prediction
and visualization of host - pathogen protein -protein interactions”,
Briefings in Bio informatics23(3),bbac 125,2022.

[13] Guan wang, yusun, Jianxin wang, “Automatic image- based plant
disease severity estimation using deep learning” computational
intelligence and neuroscience, 2017.

[14] Rajkumar, Neha Shukla, “Plant disease detection and crop

Recommendation using CNN and machine learning”,2022 international
mobile and embedded technology conference (MECON),168-172,2022.

[15] Balakrishnan Natarajan, Pushpalatha Krishnan, Contrast

Enhancement Based Image Detection Using Edge Preserved Key Pixel
Point Filtering, 42(2),423-438,2022.

979-8-3503-7519-0/24/$31.00 ©2024 IEEE 648

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY JALANDAR. Downloaded on February 01,2025 at 07:39:15 UTC from IEEE Xplore. Restrictions apply.