NORTH SOUTH UNIVERSITY

Department of Electrical and Computer Engineering

DIGITAL KIRISI: AI-Based Agriculture

Syeda Mahin Binta Haque 2012411642

Ahmed Kiser 2021280042

Israt Jahan Mitu 2022416642

Instructor: Md. Ishan Arefin Hossain (IAH)

CSE 299: Junior Design

Section 19

Team FrontERA
 Table of content

Abstract 1

Introduction 2

Literature review 4

Motivation 6

Aim 7

Application of the project in real life 8

Tools 10

Projected plan 11

Projected timeline 13

Projected Cost 13

Conclusion 14

Contribution 14

References 15
 List of Figures

Aim 2

Purpose 3

Prediction Block Diagram 4

Classification Block Diagram 5

Basic Functionality 6

Classification System Architecture 12

 List of Tables

Need and Opportunity 6

Tools 10
Timeline 13
Abstract:

Bangladesh is predominantly an agricultural country where agriculture is vital in accelerating economic

growth. The population is rapidly expanding, and with this growth, the demand for food is also increasing.
Farmers must use toxic pesticides more frequently and intensively on the soil to meet the growing demand
because traditional methods are insufficient. Moreover, this practice can occasionally result in crop failure,
wasting a lot of crops, money, and labor. To meet the demands, new automated technologies known as AI
were established which brought about an agricultural revolution, with the use of this method all of our
traditional problems can be handled in the blink of an eye. Our DIGITAL KIRISHI is an AI-based
comprehensive solution where we are using machine learning techniques to predict crop yields, detect
disease in crops, and optimize irrigation and fertilization. Since other existing solutions allude to pricey
IOT-based support and systems that are fully automated for which our farmers may lose their jobs, farmers
in our system can do their jobs precisely without the worry of losing them.

Keywords—Machine Learning, Deep Learning, Leaf Disease Detection, Image Classification

1
Introduction:

As an agriculture-based country with more than 40% of the workforce working in agriculture, we all know
that agriculture is important for people's livelihood, employment, and GDP contribution. Its contribution
has decreased over the last decade, falling from 17.6% in 2010 to 12.6% in 2020. Agriculture faces many
challenges on a daily basis and is not a smooth-running business one of the main reasons for this decline is
our traditional farming method, our farmers are not sufficiently educated to make precise decisions about
which crop is suitable for which land quantity of pesticides and fertilizers and also they can't determine the
disease for which they can't properly take care of their land, resulting in them being unable to meet
consumer demand and losing a lot of crops also money With the continued rise of the human population,
the demand for the agriculture industry will increase, and precision farming has become increasingly
important in today's globe.

Our AI-based project has the potential to revolutionize modern agriculture. This project utilizes the power
of machine learning to predict crop yield, detect diseases in crops, and optimize irrigation and fertilization
techniques, leading to more efficient and sustainable farming practices. We are using machine learning
algorithms for large datasets of environmental, climate, and crop-related data to predict which crops are
likely to grow well in a particular area, detect crop diseases, and optimize irrigation and fertilization
practices for the best possible crop yields.

The initiative aims to assist farmers in increasing crop productivity by making accurate predictions of the
best crops to plant based on environmental characteristics such as soil type, climate, and weather patterns.
The research can detect which crops are most likely to thrive in a certain area by using machine learning
algorithms to analyze massive volumes of data, saving farmers time and costs.

The research is capable of detecting crop illnesses as well as predicting crop yield. The system can recognize
indicators of disease in real time by evaluating crop photos and applying neural network. These models are
meant to learn and extract crucial information from the images that enable them to classify the crops
appropriately. Furthermore, the project can optimize irrigation and fertilization techniques, which can save
farmers water and reduce the amount of fertilizer needed to grow healthy crops. By analyzing data on soil
moisture levels, weather patterns, and other factors, the system can determine the most efficient irrigation
and fertilization schedule, leading to healthier plants and reduced costs.

2
Machine learning (ML): It is an area of AI that includes teaching robots to discover patterns from data
without being explicitly programmed. Machine learning can be used in agriculture to produce predictions
and improve decision-making by analyzing data on parameters such as weather, soil conditions, and crop
yields.

Neural Network: The structure and function of the human brain-inspired neural networks, a form of the
machine learning algorithm. They are made up of interconnected nodes, or neurons, that process and
transfer data. Neural networks can be trained on data to create predictions or classifications, and they can
be used to assess data from sensors, drones, and other sources in agricultural systems.

CNN: Convolutional neural networks (CNNs) are a type of neural network that is often employed in image
recognition tasks such as object recognition in pictures. CNNs can be used in agriculture to evaluate
photographs of crops or fields, such as recognizing weeds or diseases. CNNs can also be used to produce
predictions or classifications from other forms of data, such as time series data.

Linear Regression: A variable's value can be predicted using linear regression analysis based on the value
of another variable. The association between many environmental factors like temperature, rainfall, etc. and
agricultural yield is determined by using linear regression. It is crucial to estimate the potential crop
production rate because the farmers will profit from the outcome of this forecast. This algorithm is the best
to apply when you know the relationship between the independent and dependent variable has a linear
relationship because it is less difficult than other algorithms.

3
Literature review:

The author of identifying tomato leaf diseases by using a convocation neural network. The author divided
their work into three major phases. First, they prepare images of 18160 images belonging to 10 different
classes. From that, 4800 images were for testing and 13360 images were used for training. Images are in
unstructured order and JPG format. Then they pre-process data by resizing 60 X 60 resolution. Then use a
convolutional neural network to classify the dataset. Here convolutional neural networks (CNN) are used
for creating the computational model, which works on those unstructured image inputs and converts them
to corresponding classification output labels. The model performs 94.8% accuracy in 30 epochs and
reported 99.3% of training accuracy. So, an average of 94% accuracy is obtained.[1]

The author user Support Vector Machines in Rice Crop Yield Prediction. The author uses the SVM
algorithm to observe the binary classification [2]. There is root relative squared error of 82.51% in that
model. According to the result, other algorithm performs better than SVM.[2]

The model accurately predicts the plant or flower when any random image is fed into the system. By a
softmax classifier. In VGG16, ‘16’ stands for the number of weight layers in the network. Keras library in
python includes VGG, ResNet, Inception, and Xception network architectures The model represented
perfectly identifies and gives an accuracy of up to 99.53%.[3]

This method has varying effectiveness mostly between 52 and 75% for discriminating between the plants
and weeds. They utilize four interfere with wellsprings of P80C592 in the framework, which are the clock
intrude on the source, the outer intrudes on the source, the ADC end-of-transformation intrude on the source
and the UART sequential I/O port intrude on the source.[4]

A journal published by Australia in 2018, The topic is machine learning-based techniques for accurate crop
yield prediction and nitrogen status estimation. The discussion of this topic is about calculating Nitrogen
requirements and ML techniques which are Artificial Neural Networks, Support Vector Regression, M5-
Prime Regression Trees, and k-nearest neighbor. Applications of SVMs include crop type classification,
Artificial Neural Networks (ANNs) are widely used in remote sensing to predict crop yield, M5-Prime

4
Regression Trees are a suitable tool for multi-class crop prediction, Least Squares Support Vector Machine
is a promising tool for regression analysis to quantify Nitrogen status.[5]

A journal published by Saketh Jeedigunta, Excellencia Junior College, Hyderabad, in 2022 where told that
Machine learning algorithms such as SVM, DT, and ANN can be used for forecasting to predict agricultural
production. With 99.37 per cent accuracy,crop recommendation is most effective with Random Forest and
98.80% accuracy, Random Forest delivers the best recommendations for irrigation. Also, there included
the confusion matrix and Various factors like type of soil, fertilizer type, irrigation and nutrient content,
and weather condition. By forecasting production in the early stage, farmers should take preventive
measures to improve production. The advantage of farmers by this technique is that alternative crop
production can be possible in an early stage.[6]

An article published by China in 2023 where talked about a few unbalanced raw samples of rice leaf
diseases images grouped into four diseases namely; bacterial blight(BB),tungro(TG), brown spot (BS), and
rice-blast (RB) with 1584, 1308, 1440, and 1600 images, respectively[7]. Agriculture is seriously threatened
by the lack of an accurate disease detection system and here the solution is computer vision (CV) which
made it possible to detect plant leaf disease using machine learning (ML) models. But the date is limited
here and also see that CNN is better than Gan.

5
Motivation:

Bangladesh is primarily an agricultural nation and this sector's contribution is essential to our economic
growth. Producing high-quality and quantity crops is necessary, but our farmers need help in this sector.
Farmers face difficulties in making informed decisions due to the lack of data-driven insights and
recommendations in their farming operations. The lack of proper detection of diseases and fertilizer
information makes wrong decisions for farmers. Traditional farming methods could be more efficient and
effective, leading to low productivity and profitability. Solving these types of problems takes time and
effort.

Digital Krishi represents an automated, comprehensive solution system where crop prediction and detection
of disease, optimization of irrigation and fertilization will be provided. This system will help farmers make
proper decisions without wasting time and money. We will create this system to remove the difficulties
which the farmers face.

Need Opportunity

Precise agricultural: Improved decision- Creation of employment: It will provide new jobs for
making for farmers. unemployment people.

Supply chain: Create a connection Optimization of farming operations: It will help farmers for
between producer and the consumer. decision making.

Crop Monitoring: Detect disease and Increased efficiency productivity and reduce cost: Create
give solution. sustainable growth of production with reduce costing.

Table 1: Need and Opportunity

6
Aim:

Digital Krishi is a complete solution for crop production and management. We will create a web-based
system to help farmers improve their crop production. It provides the best information on Crop Prediction,
Leaf Diseases.

Crop prediction help to decide what to

Improved decision-making for farmers. grow and Create sustainable growth of
production with reduced costing.

Digital Krisi

It will provide new jobs for unemployed

Detect leaf disease and give the solution.
people.

Fig.1: Aim of our project

Automating Farming
Operation

Improving Productivity

Purpose
Improved resource
management

Increased efficiency

Fig.2: Purpose of our project

7
Application:

Digital Krishi, this system will be created for the solution of many farmers' problems. This web-based
solution system has some features like Crop Prediction, and Detect Leaf Disease. Farmers are not user-
friendly with technology. So, this project will also include a team of professionals who will educate farmers
on these technologies and the features of this system.

Crop Prediction: In this feature, there will be a form which has to fill up by the user.

Predict of Crop

Nitrogen:

Phosphorus:
After fill up this form it will give proper
prediction of crop. By sowing the right seeds, the
Potassium: farmer can get good yield in his land.

PH Level:

Predict

Fig.3: Prediction Block Diagram

8
Detection of Leaf Diseases: In this feature, there will be an option for upload.

Detect Disease
Please upload the image: After uploading the image file, it will
Choose File detect the disease and give instant solution
to farmer.
Predict

Fig.4: Classification Block Diagram

9
Tools needed and short description:

As we are building a web application implementing machine learning, we can divide our project into
three-part, Frontend, Backend and Machine learning models.

Frontend Backend Machine Learning Model

Tools: Tools: Tools:
Language: HTML, CSS, Language: Python Language: Python
JavaScript Framework: Django Framework/library: NumPy,
Framework: Bootstrap, React Pandas, Seaborn, Matplotlib,
JS, Next JS PyTorch

We will use Python as a

We will use HTML for the Our web application will contain
programming language to build
skeleton structure and CSS for Django as the backend.
a machine-learning model.
the basic design. And bootstrap
for responsive design. Besides After building the machine
We will use PyTorch to develop
those, we will use React Js and learning model, we will connect
the neural network model.
Next JS for the front end. those models with Django.

Table 2: Tools

Also, we will use google translate API to translate our website. For version control, we will use GitHub,
and for the design prototype, we will use Figma. And we will use the SQLite database. We will use google
collab to build a machine-learning model.

10
Project Plan:

The basic functionality of our web project is, this project could be used to analyze the collected data and
make predictions. This could include predicting crop yields, detecting disease in crops, or optimizing
irrigation and fertilization and based on the analyzed data, the project could provide farmers with
recommendations and best practices for their operation and also the project could detect disease in crops.

Data analysis

Basic Decision Support

Functioality

Diseases detection

Fig.5: Basic Functionality

11
First, we have to collect a dataset for our project. After that, we have to divide those data into test and
training. Then using torchvision we will load and normalize test and training data. Then we have to build
a convolutional neural network. Then we have to define the lost function and train the model. Finally we
can test the model by testing data.

Load and
normalize the
training and test
datasets

Define Neural
Network

Define loss
function

Train the network

Test the network

Fig.6: Classification System Architecture

12
Projected timeline:

Work Approximate time

UI design 1 week

Registration and log in 2 weeks

Prediction Model 3 weeks

Image classification Model 5 weeks

Connect model with backend 1 week

Report writing 1 week

Table 3: Timeline

Projected Cost:

We don’t apply any hardware tools so we don’t have any project cost. But for training the deep learning
model we required GPU.

13
Conclusion:
According to a recent study, total household food waste in Bangladesh is roughly 10.62 million tons per
year, which could be a significant concern if not effectively managed. To solve this crisis AI took a
revolutionary place. This project enables farmers to make data-driven decisions that boost crop yields, cut
pesticide and fertilizer use, and reduce water consumption by utilizing the power of AI and machine
learning. we can efficiently solve the problem with a low budget by saving farmers' jobs.

Overall, we can say this project has the potential to transform the agricultural industry by providing farmers
with powerful tools that increase efficiency, reduce waste, and maximize profits.

Contribution:

1. Syeda Mahin Binta Haque

• UI design
• Register and Log in
• Data collection
• Classification

2. Ahmed Kiser
• Implement crop prediction
• Image classification
• Connect model in backend
• Data Process

3. Israt Jahan Mitu

• Database manage
• Responsive design
• Data collection
• Prediction Model build

14
References:

[1] P. Tm, A. Pranathi, K. SaiAshritha, N. B. Chittaragi and S. G. Koolagudi, "Tomato Leaf Disease
Detection Using Convolutional Neural Networks," 2018 Eleventh International Conference on
Contemporary Computing (IC3), Noida, India, 2018, pp. 1-5, doi: 10.1109/IC3.2018.8530532.

[2] N. Gandhi, L. J. Armstrong, O. Petkar and A. K. Tripathy, "Rice crop yield prediction in India using
support vector machines," 2016 13th International Joint Conference on Computer Science and Software
Engineering (JCSSE), Khon Kaen, Thailand, 2016, pp. 1-5, doi: 10.1109/JCSSE.2016.7748856.

[3] K. Jha, A. Doshi, P. Patel, and M. Shah, “A comprehensive review on automation in agriculture using
artificial intelligence,” Artificial Intelligence in Agriculture, vol. 2, pp. 1–12, Jun. 2019, doi:
https://doi.org/10.1016/j.aiia.2019.05.004.

[4] T. Talaviya, D. Shah, N. Patel, H. Yagnik, and M. Shah, “Implementation of artificial intelligence in
agriculture for optimisation of irrigation and application of pesticides and herbicides,” Artificial
Intelligence in Agriculture, vol. 4, Apr. 2020, doi: https://doi.org/10.1016/j.aiia.2020.04.002.

[5] A. Chlingaryan, S. Sukkarieh, and B. Whelan, “Machine learning approaches for crop yield prediction
and nitrogen status estimation in precision agriculture: A review,” Computers and Electronics in
Agriculture, vol. 151, pp. 61–69, Aug. 2018, doi: https://doi.org/10.1016/j.compag.2018.05.012.

[6] S. Jeedigunta, “Using Machine Learning Technique for Recommendation of Fertilizer and Irrigation
in Crops,” International Journal of Technology, Science and Engineering, vol. 5, pp. 2457–1016, 2022,
Accessed: Feb. 21, 2023. [Online]. Available:
http://www.bharatpublication.com/admin/upload_ijtse/03%20Saketh%20Jeedigunta%2001298.pdf

[7] Y. Haruna, S. Qin, and M. J. Mbyamm Kiki, “An Improved Approach to Detection of Rice Leaf
Disease with GAN-Based Data Augmentation Pipeline,” Applied Sciences, vol. 13, no. 3, p. 1346, Jan.
2023, doi: https://doi.org/10.3390/app13031346.

15