M-Velanmai Application: Leveraging AI Based Extension Advisory

System for Rice Farmers

Karthikeyan Chandrasekaran

, Shri Rangasami S. R.

, Pazhanivelan S.

and Aravindh Kumar S

4* d

Department of Agricultural Extension and Rural Sociology, TNAU, Coimbatore, India

Department of Forage Crops, TNAU, Coimbatore, India

Centre for Water and Geo-Spatial Studies, TNAU, Coimbatore, India

Department of Plant Bio-Technology, TNAU, Coimbatore, India

Keywords: M-Velanmai, Rice Plant Protection, Artificial Intelligence, Machine Learning.

Abstract: “M-Velanmai” app. is an interactive, demand driven and personalized android mobile based extension

advisory system for accessing, appropriate and timely technological information / decision support in

agriculture by the farmers of Tamil Nadu. The app is designed to work on android mobile phones. “M-

Velanmai” uses the techniques of artificial intelligence and machine learning to provide two types of support

services viz., decision support and information support for the benefit of farmers. The farmer will be given

provision to record their response regarding the level of satisfaction upon adoption of advisories as a feedback

for the extension service availed. The “M-Velanmai” app. developed for rice can be extended for other major

crops of Tamil Nadu in near future.

1 INTRODUCTION

Climate change, pest and disease outbreaks, scarcity

of water, labor crisis, rising population and a resultant

rise in food demand are the challenges that agriculture

sector is dealing with. To overcome these challenges

farmers must be provided with diverse of precise

information throughout the crop period in order to

make well-informed decisions and overcome these

challenges in agriculture. Apart from these

constraints, global pandemic is also limiting farmer’s

interaction with scientists. Due to low rate of adoption

of agricultural technologies by farmers there is no

significant increase in crop productivity. This

highlights the necessity to employ highly

sophisticated technologies like AI to help farmers in

enhancing the outcomes from agriculture. Artificial

intelligence (AI) refers to the simulation of human

intelligence in machines that are programmed to think

like humans and mimic their actions (Frankenfield,

2021).

http://orcid.com/0000-0002-0631-3820

http://orcid.com/0000-0003-4804-1816

http://orcid.com/0000-0002-3596-3232

http://orcid.com/0000-0002-4185-8826

Unique advantage of applying AI in agriculture is

that objective decisions are made based on

quantitative assessment of data unlike the case of

relative and subjective decisions made by

farmers/experts based on visual examination. AI is

utilized in the field of agriculture for:

1. Increasing the share of price realization to

producers (Eg: Predictive analytics using AI tools to

get supply and demand information to famers)

2. Soil health monitoring and restoration (Use

of image recognition an DL models)

3. Crop health monitoring and providing real

time action advisories to farmers (To predict

advisories for sowing, pest control, input control)

4. Analyzing farm data (Analysis of weather

conditions, temperature, water and soil conditions)

5. Seasonal forecasting (To create seasonal

forecasting models to improve agricultural accuracy

and increase productivity)

6. Precision agriculture (NITI Aayog and

IBM have collaborated to develop an AI-based crop

116

Chandrasekaran, K., R., S. S., S., P. and Kumar S, A.

M-Velanmai Application: Leveraging AI Based Extension Advisory System for Rice Farmers.

DOI: 10.5220/0012882900004519

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Emerging Innovations for Sustainable Agriculture (ICEISA 2024), pages 116-120

ISBN: 978-989-758-714-6

yield prediction model to detect diseases, pests, and

poor plant nutrition)

7. Tackling the labor challenge (AI agriculture

bots that augment the human labour workforce)

8. Customized and personalized assistance to

farmers (AI based Chatbots to provide

recommendation and advice on farmer’s problems)

9. Increasing efficiency of farm

mechanization (Image classification tools combined

with remote and local sensed data).

M-Velanmai (‘Mobile-Agriculture’ in English) –

is an Artificial Intelligence based bilingual extension

advisory system delivered through android mobile

application is conceived to extend decision support

services. The app is being created using deep

learning/ machine learning technologies to detect the

crop damage symptoms (insects and diseases)

instantly & offer reliable and cost-effective advisories

in one of the major crop of Tamil Nadu i.e. Paddy.

(Karthikeyan, 2020)

The app has been designed to work on android

mobile phones. M-Velanmai uses the techniques of

artificial intelligence and deep learning to provide

two types of support services viz., decision support

and information support mainly for the benefit of

farmers in Tamil Nadu.

The app. will be user friendly which does most of

the operations in single touch as it has been intended

for the farmers (users) were the user is given

provision to record their response regarding the level

of satisfaction upon adoption of advisories as

feedback for the extension service availed.

2 MATERIALS AND METHODS

In order to build an AI model using Machine

Learning/ Deep Learning technologies the following

five major activities need to be carried out which

forms the basis for creating the framework in M-

Velanmai application development. They are,

A. Data collection

B. Data pre-processing

C. Data Augmentation

D. Feature extraction

E. Object Detection

A. Data collection

High quality images of the crop damage symptoms

with varied resolutions under diverse backgrounds

were captured using Digital camera and Smartphones.

The image library was built by capturing more than

80,000 pictures of damage symptoms of Paddy in real

field situations under all crop growth stages and

varied environmental conditions from the pest

hotspots of Tamil Nadu.

The data collection forms the first step in building

the pest detection model. It is well aware that, the

quality of output is deeply influenced by the quality

of input, high quality images were collected by DSLR

and smartphones under diverse scenarios such as

different resolutions, under different crop growth

stages in different times of the day i.e., morning,

afternoon and evening hours in order to better

represent realistic working conditions in various

environmental and illumination conditions. Various

studies have reported that the AI models developed

with inputs obtained from controlled conditions fail

to make accurate predictions in real field situations.

Taking this into consideration, we captured the

images in real field settings with and without

background under different lighting conditions to

improve the reliability of the performance of the

model.

B. Data Pre-Processing

Data pre-processing involves manual labeling and

cropping the images to give major focus on the region

with symptoms in the photos captured. To work with

Convolutional Neural Network (CNN) models, large

datasets with manually labeled targets are required.

Initially when it was started to train the model to

classify the infected images from the healthy ones, it

has been the issues of misclassifications with a dip in

the prediction accuracy was confronted. Then it has

been tried with cropped images where we could

observe an improvement in the model performance

metrics. Hence, data pre-processing work serves as

the mandatory prerequisite for preparing the data for

training the machine. Here, the symptoms were

labeled used LabelImg software.

C. Data Augmentation

The next step in the process of AI model development

is data augmentation. Data augmentation denotes

increasing the input which involves rotation of the

existing images randomly, zooming in, flipping the

images, etc. Random brightness was given to the

images to replicate the field level settings as the

intensity of sunlight fluctuates throughout the day

inducing changes in the color features. Hence, data

augmentation step is considered crucial assuming that

the machine will be trained for all the field conditions

and is expected to produce reliable prediction in real

time situations.

M-Velanmai Application: Leveraging AI Based Extension Advisory System for Rice Farmers

117

D. Feature Extraction

Features such as color, texture, shape were

extracted. Feature extraction is done to retain the

patterns of the damage symptoms without loss of

information. Four coordinates namely X, Y axis,

height and width were considered to identify (or)

recognize the datasets.

E. Object Detection

Lastly, CNN architectures such as VGG 16, ResNet

were trained based on the extracted features for

classification into stem borer infested and healthy

leaf. The ResNet model performed better compared

to VGG 16 with the training accuracy of 98 % and

Validation accuracy of 95 %. Further to achieve cent

per cent accuracy, we tried Object detection model

YOLO V3. Object detection refers to the

identification of each object present in the given

image. The YOLO V3 model identified the presence

of white ear in the image indicated by the boxes in the

image. The percentage of white ear present in each

box is reflected in the predicted output image. If we

take mean of the percentage of white ear in each box,

we can arrive the percentage of white ear damage

symptom present in the picture comprising of healthy

grains, leaves etc.

3 RESULTS AND DISCUSSIONS

Data collection and labeling is a tedious work. More

training samples (large datasets) are needed in order

to predict the disease more accurately and improve

the generalization. One particular CNN model will

not be the best for prediction of different symptoms

expressed in different plant parts. Speed of detection

needs to be faster which requires high computational/

IT infrastructure resources like GPU, Tensorflow.

The framework architecture of M-Velanmai

extension advisory system comprised of three major

components. They are, 1. End users 2. Features and 3.

Database.

1. End users:

The End users are farmers, extension workers and

agricultural professionals who might seek extension

advisories through the M-Velanmai application,

presently designed for paddy crop.

User friendly interface has been created for the access

by users through two platforms namely an android

mobile application and web application so that the

project administrators and agricultural experts

(TNAU Scientists) can use both the mobile and

personal computer devices to access M-Velanmai

system. However, farmers (users) access is restricted

to mobile application only. The users are allowed to

access the M-Velanmai system through registration

process authenticated by the application

administrators (TNAU).

2. Features:

The computational features of the M-Velanmai

system consists of eight units namely,

1. Personal Settings

2. Personal Analytics

3. Analytics dashboard

4. Recommendation Engine Editor

5. Report template editor

6. Deep learning Engine

7. Rule based AI

8. Report Generation

Once the user hits the M-Velanmai system, the 1.

Personal Settings, 2. Personal Analytics and 3.

Analytics dashboard gets activated to acquire and

process the input data provided by the users.

If the system receives image of query as an input

data, the Deep Learning engine gets activated to

identify the pest image and recognise it using Rule

based AI approach. In case, the AI model is able to

predict the input (image) to an extent of >95 percent,

then the appropriate advisory for tackling the pest

problem/query raised by the farmer in the form of an

input will be generated by the system (Report

generation).

3. Database:

The backend Database of M-Velanmai consists of

Images, Machine Learning models, Algo Hyper

Parameters, Farmer’s data and Knowledge bank.

These are stored in the cloud server to access

recognise and deliver the appropriate solutions based

on target query received from the users of the android

application.

The database component of M-Velanmai architecture

consists of five units namely Images, Machine

Learning models, Algo Hyper Parameters, Farmer’s

data and knowledge bank. The farmer/ user’s data and

input data (image/text) regarding the images of the

symptoms collected from field (approximately 5000

per symptom) and the advisories about the package of

practices, management of pest/ diseases/ nutritional

disorders etc., of a paddy crop form a part of the

knowledge bank. The suitable algorithms and AI

models/ CNN models fit to predict the pest image

provided as input are also stored under Machine

Learning models other weather-based parameters,

environmental parameters to be considered at the time

of pest incidence in farmers’ field are stored under

ICEISA 2024 - International Conference on ‘Emerging Innovations for Sustainable Agriculture: Leveraging the potential of Digital

Innovations by the Farmers, Agri-tech Startups and Agribusiness Enterprises in Agricu

118

Algo hyper parameters unit which would also be

considered to deliver extension advisories to the users

of M-Velanmai system.

Figure 1: Technology transfer process automated in M-

Velanmai application

Figure 2: The work flow of the technology outreach process

involved in M-Velanmai – Extension model

Features of M-Velanmai app.

The M-Velanmai application is designed for use by

both farmers and agricultural experts in interactive

mode. The special features incorporated in the

android application are as follows (Karthikeyan,

2021).

• Bilingual Mode of operation

• Query registration in the various forms such as

images, voice, text, etc.

• Instant crop protection advisory to farmers based

on Artificial Intelligence

• AI inference validation and Query redressal by

TNAU scientists

• Conversational platform supported by text and

voice input between farmers & TNAU scientists,

• Continuous flow of push notifications on daily/

weekly basis for crop monitoring,

• Guidance on crop cultivation practices

• Send risk alerts to farmers in case of pest

outbreak.

Benefits of M-Velanmai system:

• It is expected that M-Velanmai android

application developed using Artificial

Intelligence technology will deliver

appropriate technical advisories in paddy

crop to the needed farmers instantly without

interference of human elements.

• Farmers can get personalized solutions for

their pest related problems instantly in the

form of text / voice message in English and/

or Tamil language.

• M-Velanmai ensures two-way

communication between farmers and

agricultural experts and within farmers

through creation of an interactive platform

to share / exchange information among the

users.

• M-Velanmai will serve as technology

transfer mechanism to address farmers’

problems as it saves time and cost of farmers

in reaching the experts.

• M-Velanmai will cater to the dynamic

technological, market and weather-based

information needs of farmers.

Farmer

upload the

picture of

suspected

insect

attack on

crop or

unusual

developme

nt of crop

texture

which may

be nutrient

deficiency/

pest/disease

/weeds.

While

uploading

the picture

farmer

selects the

crop

AI/ ML

solution

based

engine

scans the

images

with

similar

patterns

from its

data base

on various

algorithms

predicts

the

probable

in the

crop.

The result of

the problem

is viewed to

the farmer

along with

remedial

measures

and advisory

mechanism

with

possible

work flow in

between

wherein

Agriculture

department

official can

edit the

system

generated

advisory

M-Velanmai Application: Leveraging AI Based Extension Advisory System for Rice Farmers

119

Figure 3: M-Velanmai Menu Page.

4 CONCLUSION

Decision making in the future will be a complex mix

of human and computer factors as farmers are seeking

ways to improve profitability and efficiency. AI

based pest detection smartphone applications can

support farmers in monitoring the crop’s health by

identifying the pests in the field at an early stage.

More pragmatic farming can take place with the

support of AI which helps in improving agricultural

yield and reduce potential environmental risk.

However, it is crucial to test and validate the

emerging AI applications in agriculture sector as it is

impacted by uncontrolled environmental factors

unlike other sectors where risk is easier to model and

predict. Hence, this AI powered M-Velanmai

application would serve as a technology provider to

the farmers besides an eye opener for agricultural

scientists to contribute towards more systematic

research on AI in agriculture.

5 CONFLICT OF INTEREST

There is no conflict of interest.

ACKNOWLEDGEMENT

The authors acknowledge the support rendered by

World Bank Aided TNIAM Project to fund the study.

Technical guidance provided by FarmWise AI

consultants are acknowledged.

REFERENCES

Frankenfield, J. (2021). Artificial Intelligence (AI).

Retrieved from

https://www.investopedia.com/terms/a/artificial-

intelligence-ai.asp

Karthikeyan, C. 2020. M-Velanmai – Artificial Intelligence

Based Agricultural Extension System. In the Souvenir

and Extended Abstracts of the International Conference

on Recent Trends in Agriculture Towards Food

Security and Rural Livelihood: 17-19p.

Karthikeyan, C. 2021. Innovative Approach for Automated

Extension Advisory Services. In the Book of Abstracts

of the International Conference on Changing

Perspectives in Agricultural and Horticultural Research

for Sustainable Development, AIASA & Annamalai

University: 4-5pp.

‘Notificati

on icon’-

To check

the recent

notification

received

‘Menu

icon’-

Profile &

language of

advisory

can be

Solved and

Unsolved

enquiries

were listed

Identifies the

pest and

provides

recommenda

tion

accordingly

Gives

informatio

n regarding

the

different

cultivation

aspects.

Delivers

advisories

based on

your

sowing date

Latest

nearby

market

price of

different

crops can

be checked

Weather

advisories

from

TAWN,

IMD and

Meghdoot

Feedback

regarding

advisories

can be

mentioned

ICEISA 2024 - International Conference on ‘Emerging Innovations for Sustainable Agriculture: Leveraging the potential of Digital

Innovations by the Farmers, Agri-tech Startups and Agribusiness Enterprises in Agricu

120