Plant Disease Identification and Pesticides Recommendations Using

CNN Deep Learning

Jujaray Vyshnavi, Sangapatnam Sowmya, P. Dhanvitha,

Mandozai Ayesha Khatoon and L. Sandhya Rekha

Department of Computer Science and Engineering (CSE‑AI), Ravindra College of Engineering for Women, Kurnool,

Andhra Pradesh, India

Keywords: Plant Disease Detection, Deep Learning, CNN, Pesticide Recommendation, Precision Agriculture.

Abstract: Diseases on plants are a major threat to global agricultural productivity and cause great economic losses and

problems of food security. Current disease detection methods employ manual inspection, which is slow,

mistakes prone and aims to disease specific knowledge. Therefore, to mitigate these challenges, there is need

to develop an AI driven Plant disease identification and pesticide recommendation system using convolutional

neural networks (CNNs). Using deep learning techniques, the leaf images are automatically classified to

generate plant disease to a very high accuracy. Thereafter, it offers real-time and disease dependent pesticide

recommendations, maximizing treatment efficiency while reducing pesticides not needed. The model is

trained on the available diseased and healthy plants images after a few preprocessing operations, then feature

extraction and classification using CNN architecture. One main benefit of this method is the availability for

real time disease diagnosis, the decreased dependence of agricultural experts, increased crop yield and usage

of environmentally sustainable pesticides. The system is web or mobile application deployable so it can be

distributed to the farmers. Also, we can further improve the predictive accuracy by tracking environmental

condition like temperature, humidity, and soil health through integrating IoT. The purpose of this research is

to provide an automated, scalable, and cost-effective solution to the problem of plant disease management in

order to enhance precision agriculture. Although this existing system is useful for the future, it may be further

improved this way: multi disease detection, disease affected zone localization, and cloud-based updating for

continuous learning.

1 INTRODUCTION

Agriculture as a sector is an important sector to assure

food security and stability of the economy globally.

In this way, however, plant diseases are preventing

crop production and significantly decreasing yields,

and ending up farmers frustrated, malleating millions

of dollars to them. Detecting plant diseases

traditionally has been a manual raster that requires

expert inspection and is laborious, costly and quite

often inaccurate. If timely detection is not possible, it

can lead to infections that spread throughout the

place, causing harm to agriculture and food supply.

As Artificial Intelligence (AI) and Deep Learning

(DL) technologies have recently advanced, it has

become a promising approach of automated disease

detection. Deep learning model called Convolutional

Neural Networks (CNNs) are shown to be very

effective for image classification and pattern

recognition and therefore are suitable for leaf image-

based plant disease diagnosis. CNNs have the ability

to learn intricate features from plant images to spot

slight disease symptoms that may not show so clear

to human eye.

In this study, there is a suggestion of an AI

assisted Plant Disease Identification and Pesticide

Recommendation System based on CNNs for real

time disease classification. The system takes plant

leaves images, detects the possible diseases, and then

provides the most adequate pesticides for treatment.

As an automated system, it allows the reduction of

dependency on agricultural specialists as well as

quick and accurate detection of a disease to farmers.

Sustainable pesticide management is one of the

most important advantages of this procedure.

Inappropriate and excessive use of pesticides can be

harmful to the environment, lead to pesticide

resistance and raise the cost to farmers.

Vyshnavi, J., Sowmya, S., Dhanvitha, P., Khatoon, M. A. and Rekha, L. S.

Plant Disease Identiﬁcation and Pesticides Recommendations Using CNN Deep Learning.

DOI: 10.5220/0013909500004919

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

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 4, pages

153-160

ISBN: 978-989-758-777-1

153

Recommendation of most appropriate pesticide

according to diagnosed disease is given by the system

such that pest control is efficient with less chemical

exposure, leading to eco-friendly farming.

This system has been implemented via mobile and

web application so that farmers in most remote areas

can easily access to this system. Future ways of

enhancement can be real time IoT based monitoring,

multi disease detection, and cloud-based update to

keep on enhancing the accuracy. This dissertation is

targeted to revolutionize the Precision Agriculture, by

integrating the AI and deep learning in order to

improve the crop health, productivity, and

sustainability.

2 RESEARCH METHODOLOGY

The research methodology of Plant Disease

Identification and Pesticide Recommendation System

using CNN is implemented following a structured

approach so that diseases can be detected accurately

and pesticides can be recommended effectively. The

steps included in the methodology are:

2.1 Data Collection and Preprocessing

• A large set of plant leaf images are gathered

from public sources like PlantVillage, Kaggle,

and agricultural research institutes.

• A variety of plant species and disease types are

categorized into healthy and diseased leaves

images.

• Image resizing, noise reduction, contrast

enhancement and data augmentation as

preprocessing techniques make the model

robust and improve the chances of accuracy.

2.2 CNN Model Development

It designs and train a Convolutional Neural Network

(CNN) to classify plant diseases.

• The model includes multiple convolutional

layers, pooling layers, and fully connected

layers for the purpose of extraction of

meaningful features from images.

• The dataset is divided into training, validating

and testing sets; and the model is trained by

minimizing classification errors via Adam

optimizer or Stochastic Gradient Descent

(SGD).

• The model performance is evaluated in terms of

performance metrics like accuracy, precision,

recall, and F1-score.

2.3 Pesticide Recommendation System

• Using expert knowledge, agricultural reports

and guidelines on pesticide usage, a disease to

pesticide mapping database is created.

• Based on the already detected disease, the

system fetches the most appropriate pesticide

recommendations based on factors such as

effectiveness, ecology friendliness and

government regulations

2.4 System Deployment and User

Interface

• It is deployed as web based or mobile

application to detect disease in real time using

trained model.

• To mention, farmers can also upload plant leaf

images of a farm and the system gives instant

disease identification which can also suggest

pesticides.

• Methods of combining cloud storage and IoT

integration for real time monitoring of the plant

health or the surrounding environment is

discussed.

2.5 Testing and Validation

• Practical effectiveness of the system is tested

using real world plant images.

• The accuracy, efficiency and usabiltiy of this

paradigm is compared with existing disease

detection methods.

• The system is refined based on feedback

obtained from agricultural experts and farmers

in terms of usability

2.6 Research Area

Plant diseases play an important role in agriculture

and pose serious obstacles to global food production,

while it is of great relevance in the agriculture sector

maintaining the food production sufficient for

humanity. Currently, the joint hands with agricultural

specialists to identify these diseases include manual

inspection, which is time consuming, costly and error

prone. While purpose is logically placed in the

introduction section, another clear reader impression

is that it can also be placed in the conclusion part. In

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

154

this paper, the integration of deep learning methods,

in specific Convolutional Neural Networks (CNNs),

is applied toward revolutionizing plant disease

recognition and therapeutic technique.

This study's primary research is sited on precision

agriculture, which entails using AI and data enabled

methods to improve the technique of farming. Image

classification tasks have been made remarkably

successful using deep learning models, in particular

CNNs; thus, they are a wise choice to tackle the

problem of identifying plant diseases from leaf

images. This is done in order to reduce dependency

on human expertise for disease detection while

enhancing its accuracy and speed. The computer

vision algorithms implementation increases the

efficiency of disease classification because it can

identify even the subtle patterns, which can be pure

invisible to the traditional methods.

Second is sustainable pesticide management as

another important aspect of this research. The

excessive and improper use of pesticides can have

negative effects on the environment, contamination

of soil and water and development of pesticide

resistance in the pests. Here, this study includes a

pesticide recommendation system in which

appropriate pesticide treatment is suggested given the

identified disease. Recommendations generated by

the system for the targeted and regulated pesticide use

will be ecofriendly and cost-effective agricultural

practices. This fits with the modern concept of

sustainable farming: to decrease chemical overuse

and to increase crop health.

In addition, the research is also extended to IoT-

based smart farming by incorporating real time

monitoring systems. Environmental factors such as

temperature, humidity, and soil conditions which

would affect disease outbreaks can be tracked by

sensors. The system, by being based on an IoT

technology, can be enhanced through the

incorporation of AI driven disease detection

simultaneously, and can provide predictive insights to

farmers in order to take preventive measures before

the disease spreads widely. Data storage can be cloud

based and update on real time to continuously

improve and adapt the model to new diseases.

This research is poised to make a great innovation

in modern farming, being an interdisciplinary

research at the intersection of AI, agriculture,

computer vision, and environment sustainability.

This study proposes to improve the crop health

management and ensure food security through the

synergism of deep learning with smart sensors and

data processing in real time. There are potential

outcomes of this research to empower the farmers, to

enhance the agricultural efficiency and in general, to

assist a less polluting and stronger farming

ecosystem.

3 LITERATURE SYSTEM

3.1 Plant Disease Detection Using Deep

Learning

• Title: Plant Disease Identification Using

Convolutional Neural Networks.

Author: Mohanty et al.

Abstract: This study applies CNN-based

deep learning models for automatic plant

disease identification using images. The

model was trained on the PlantVillage

dataset, achieving high accuracy in detecting

multiple plant diseases. The study highlights

the advantages of deep learning over

traditional feature extraction methods and

demonstrates the effectiveness of CNNs in

real-time agricultural applications.

3.2 A Deep Learning-Based Approach

for Agricultural Disease Detection

• Title: Deep Learning-Based Plant Disease

Recognition for Smart Agriculture

Author: Ferentinos et al.

Abstract: This research focuses on using

pretrained CNN architectures such as

AlexNet, VGG16, and ResNet for plant

disease classification. The study emphasizes

the importance of transfer learning to

improve detection accuracy and reduce

computational costs. The results

demonstrate that CNN-based models can

outperform traditional machine learning

techniques like SVM and decision trees.

3.3 Smart Agriculture and IoT-Based

Monitoring for Disease Prediction

• Title: IoT-Based Smart Farming System for

Disease Detection.

• Author: Zhang et al.

• Abstract: This paper explores the

integration of IoT and deep learning for real-

time monitoring of plant health. The system

uses environmental sensors to collect

temperature, humidity, and soil moisture

data, which are analyzed alongside leaf

Plant Disease Identiﬁcation and Pesticides Recommendations Using CNN Deep Learning

155

images using CNNs. The study

demonstrates that combining IoT with AI

can provide early warning systems for

farmers, helping prevent large-scale crop

losses.

3.4 Sustainable Pesticide Management

Using AI-Based Decision Support

Systems

• Title: AI- Driven Pesticide Recommendat-

ion System for Precision Agriculture.

• Author: Kumar et al.

• Abstract: This research introduces a data-

driven pesticide recommendation system

that analyzes disease symptoms and

suggests suitable pesticides. By integrating

plant pathology databases and regulatory

guidelines, the system ensures optimal

pesticide use while minimizing

environmental impact. The study

underscores the potential of AI in reducing

excessive chemical application and

promoting sustainable farming practices.

3.5 Comparative Study of Image

Processing and Deep Learning in

Agriculture

• Title: Traditional Image Processing vs.

Deep Learning for Plant Disease

Classification.

Author: Singh et al.

Abstract: The paper compares traditional

image processing techniques (such as color-

based segmentation and feature extraction)

with deep learning models for plant disease

classification. The results indicate that

CNNs outperform traditional methods in

accuracy, scalability, and adaptability,

making them more suitable for real-world

agricultural applications.

3.6 Key Takeaways from Literature

Survey

• Plant disease detection models based on

deep learning’s CNN models achieve a very

high accuracy, compared to traditional

image processing and machine learning

techniques.

• The combination of IoT with AI based

disease detection gives the farmers real time

monitoring and warnings of early detection

of such diseases that would have meant

major crop losses.

• Pre-trained models (VGG16 and ResNet)

are used in transfer learning for achieving

better classification performance with a

smaller number of training examples.

• The use of AI based pesticide

recommendation systems help in optimizing

the use of chemicals and thus adhering

towards sustainable and eco-friendly

agriculture. Nevertheless, deep learning

presents challenges including high demand

in computation, high dependency on dataset,

and real time deployment which require

more research.

4 EXISTING SYSTEM

Currently, farmers and agricultural experts manually

identify plant diseases and recommend pesticides. In

this system, farmers not only can view the symptoms

of the disease, such as leaf discoloration, spots or

wilting, but also judge the disease and then choose a

pesticide according to experience. But, there are some

deficiencies with this method, which restricts its use

on a large scale.

Therefore, instead, some of the machine learning

(ML) and image processing techniques which are

used for plant disease detection currently need

manual feature extraction which reduces the

precision. Existing methods for plant disease

classification in conventional image processing

process utilize color base segmentation, edge

detection, and feature matching. Performance of these

methods is restricted in terms of complexity of the

plant symptoms, as well as varying lighting

conditions and different angles of leaf images.

A number of agricultural advisory systems make

pesticide recommendations that are based either on

pre-defined datasets or on expert inputs.

Nevertheless, these systems are not adaptive in the

real time and they commonly suggest the application

of pesticide with no consideration to the real time

conditions such as humidity, temperature, and soil

health. On another level, they do not tackle overuse

or misuse of chemical pesticides that results in

degradation of the environmental and heightened

chemical resistance in plants.

To support smart farming, there are some

IoTbased smart farming solutions which are

accompanied by sensors monitoring soil moisture,

temperature and humidity. Although these systems

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

156

address the problems of crop health monitoring and

management of irrigation, they lack in disease

identification and provision of disease treatment

recommendations. The prediction of disease via

cloud-based AI models is still in its infancy stage and

not fully adopted as of yet because of high

computational cost and connectivity problems in rural

areas.

Limitations of the Existing System

• Manual disease identification is error-

prone and time-consuming.

• Traditional image processing techniques

have low accuracy in complex disease

classification.

• Limited real-time adaptability in existing

pesticide recommendation systems.

• Overuse or incorrect use of pesticides,

harming the environment and soil health.

• IoT-based solutions focus mainly on soil

monitoring rather than disease detection

and treatment.

Due to these challenges, there is a growing need

for an AI-powered, automated plant disease

identification system that can provide real-time,

accurate, and sustainable pesticide recommendations

to farmers.

5 PROPOSED SYSTEM

In this approach, we introduce such an AI driven,

automated plant disease detection and pesticide

advisory system that is built on top of Convolutional

Neural Networks (CNN), IoT sensors and cloud-

based decision making. Traditionally available to the

farmers are the generic solutions i.e. manual disease

identification and pesticide application, which comes

with certain limitations and this system works

towards the advancement of the same by providing

real time, accurate and sustainable solutions. It

enables detection of plant disease well in advance and

allows particularly efficient applications of the right

pesticides by using image processing, deep learning

and smart agriculture technologies.

Figure 1 shows the architecture diagram. The

system builds a deep learning model based on

convolutional neural network (CNN) trained on the

dataset of diseased and healthy plant images who are

the core component of the system. A mobile

application or a camera module may be used by the

farmers to capture an image of a plant leaf which will

be analyzed by the system and the diseased plant leaf

will be immediately diagnosed by classifying the

disease. Without manual feature extraction, the CNN

model speeds up and adds more accuracy to the

process. However, unlike traditional machine

learning methods, CNS distinguish different complex

patterns, different colours, and different types of leaf

textures, and thus can help increase disease

classification accuracy.

Figure 1: Architecture diagram.

Besides disease detection through an image-based

system, the system also equipped with IoT sensors to

evaluate environmental factors like temperature,

humidity and soil moisture that also affect the plant

health. A cloud-based AI model analyzes realtime

data to find generated from these sensors. With this

being said, the system considers both image-based

symptoms and environmental conditions to increase

the reliability of disease predictions and prevent the

outbreak from spreading. Farmers can also get alerts

and suggestions for the changing environmental

conditions.

If a disease is known, the pesticide

recommendations are based on the disease type,

severity, and environmental conditions and are

suggested by the system which uses AI. Beyond the

social factors influencing fisher behavior – both on

long and short time scales – there are informative

interactions taking place between the physical

environment and large numbers of daily decisions

made by fishers. Moreover, it offers organic ways of

Plant Disease Identiﬁcation and Pesticides Recommendations Using CNN Deep Learning

157

treatment to farmers who are embracing the use of

organic practices in farming. The system ensures the

right amount of pesticide is applied at the time it is

needed, to prevent misuse or overuse of your

chemicals, which would lead to resistance.

The system has a web and mobile user interface

with which farmers can upload images, get their

disease reports and receive pesticide treatment

recommendations to improve usability of the system.

By using a platform, real time updates, historical

disease tracking and support for multiple languages

that can be used in multiple regions are available for

use by the farmers. The system constantly gets better

by learning from new cases involving disease and

farmer feedback. An AI powered solution to Smart

Agriculture wherein this AI helps farmers to reduce

crop loss, improve yield quality and advocate the

sustainability in farming.

6 RESULTS

Figure 2: Django server running for plant disease detection

web app.

Figure 3: Login & signup flow.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

158

Figure 4: Plant disease detection – image upload flow.

Figure 2 shows the Django Server Running for

Plant Disease Detection Web App. Figure 3 shows the

Disease Detection – Image Upload Flow.

7 CONCLUSIONS

Thus, the proposed AI driven plant disease

identification and pesticide recommendation system

provides accurate, real time and automated way to

farmers for detection of plant disease and application

of right treatment. The system integrates

Convolutional Neural Networks (CNNs) to predict

disease in images, IoT sensors for environmental

monitoring, and AI driven decision making which

helps in enhancing the precision, efficiency and

sustaining in the process of agriculture.

This system improves diagnostic accuracy

compared to the traditional manual disease

identification and pesticides application, which also

reduces human error chances. With actual time iot

Information, the Plants may be monitored to

Outsmart Where and When They Are Most Liable to

Crop Lysis Caused by Environmental Factors.

Furthermore, the system aids farmers to be better

informed when taking decisions regarding pesticide

usages, in order to allow minimum possible

environmental effect while maintaining maximum

crop health.

This system has user friendly web and mobile

interface where farmers upload their plant images, get

report instantaneously and view AI based pesticide

recommendations. This enables the system to always

be up to date with the new disease case patterns and

environmental patterns and learn continuously.

Through implementing this solution, farmers will

experience increased crop production, less losses, and

better methods of pest control, increase the

productivity of agriculture and food security. A smart

approach to the way this agriculture is done is that it

is a sustainable farming that prevents the use of

excess chemical pesticides and encourages use of eco

– friendly alternatives.

This system can be further increased in the dosage

of information in the future by providing a more

significant dataset cobers more plant species and

diseases, including satellite-based monitoring, and

predictive analytics for plant diseases outbreak. It is

the adoption of such AI powered agricultural

technologies that will become a essential component

in determining the agricultural future in precision

farming and in guaranteeing efficient use of labors,

resources, money, time spent and other variables.

REFERENCES

Abbas, A., Jain, S., Gour, M., & Khanna, A. (2021).

Tomato Plant Disease Detection Using Transfer

Learning with CNN. Artificial Intelligence in

Agriculture, 5, 11-20.

https://doi.org/10.1016/j.aiia.2021.05.002

Brahimi, M., Arsenovic, M., Laraba, S., Sladojevic, S.,

Boukhalfa, K., & Moussaoui, A. (2018). Deep Learning

for Plant Diseases: Detection and Saliency Map

Visualisation. Human and Machine Learning, 93-117.

https://doi.org/10.1007/978-3-319-90403-0_5

Ferentinos, K. P. (2018). Deep Learning Models for Plant

Disease Detection and Diagnosis. Computers and

Plant Disease Identiﬁcation and Pesticides Recommendations Using CNN Deep Learning

159

Electronics in Agriculture, 145, 311-318.

https://doi.org/10.1016/j.compag.2018.01.009

Kamilaris, A., & Prenafeta-Boldú, F. X. (2018). Deep

Learning in Agriculture: A Survey. Computers and

Electronics in Agriculture, 147, 70-90.

https://doi.org/10.1016/j.compag.2018.02.016

Liu, J., Wang, X., & Wang, Z. (2020). A Comprehensive

Review on Plant Disease Detection Using Deep

Learning Techniques. Advances in Intelligent Systems

and Computing, 1021, 27-45.

Lu, J., Hu, J., Zhao, G., Mei, F., & Zhang, C. (2017). An

Improved CNN Model for Recognition of Plant

Disease. International Journal of Engineering Research

& Technology (IJERT), 6(4), 32-38.

Mohanty, S. P., Hughes, D. P., & Salathé, M. (2016). Using

Deep Learning for Image-Based Plant Disease

Detection. Frontiers in Plant Science, 7, 1419.

https://doi.org/10.3389/fpls.2016.01419

Picon, A., Alaña, M., Irigoyen, I., Irigoyen, E., &

Bereciartua, A. (2019). Deep Learning for Plant

Disease Detection with Limited Data Sets. Computers

and Electronics in Agriculture, 167, 105093.

https://doi.org/10.1016/j.compag.2019.105093

Ramcharan, A., Baranowski, K., McCloskey, P., Ahmed,

B., Legg, J., & Hughes, D. P. (2017). Deep Learning for

Image-Based Cassava Disease Detection. Frontiers in

Plant Science, 8, 1852.

https://doi.org/10.3389/fpls.2017.01852

Saleem, M. H., Potgieter, J., & Arif, K. M. (2021). Plant

Disease Classification: A Review of Deep Learning-

Based Approaches. Frontiers in Plant Science, 12,

624094. https://doi.org/10.3389/fpls.2021.624094.

Sharma, P., Kumar, S., & Shukla, A. K. (2020). Plant

Disease Identification and Classification Using CNN.

Journal of Artificial Intelligence and Machine Learning

(JAIML), 2(1), 45-57.

Sladojevic, S., Arsenovic, M., Anderla, A., Culibrk, D., &

Stefanovic, D. (2016). Deep Neural Networks Based

Recognition of Plant Diseases by Leaf Image

Classification. Computational Intelligence and

Neuroscience, 2016, 3289801.

https://doi.org/10.1155/2016/3289801

Too, E. C., Yujian, L., Njuki, S., & Yingchun, L. (2019). A

Comparative Study of Fine-Tuning Deep Learning

Models for Plant Disease Identification. Computers and

Electronics in Agriculture, 161, 272-279.

https://doi.org/10.1016/j.compag.2018.03.032

Wu, D., Lv, S., Jiang, M., Song, H., Wang, X., & Cheng,

X. (2021). Pest and Disease Detection for Smart

Agriculture Using Machine Learning Techniques.

Smart Agricultural Technology, 1, 100003.

https://doi.org/10.1016/j.atech.2021.100003

Zhang, S., Zhang, S., Zhang, C., Wang, X., & Shi, Y.

(2019). Cucumber Leaf Disease Identification with

Global Pooling Dilated Convolutional Neural Network.

Computers and Electronics in Agriculture, 162, 422-

430. https://doi.org/10.1016/j.compag.2019.04.018

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

160