Implementing an Potato

Leaf Disease Classifier

By
Rohit Hait (12020002001124),
Rajdip Banerjee (12020002001077) and
Akash Kar (12020002001050)
(B.Tech 3nd Year)
Under the Supervision of
Prof. Hriday Banerjee
Dept. of Computer Science & Engineering
University of Engineering & Management, Jaipur
Dept. of CSE, University of Engineering &
Management Jaipur
 Outlines
Introduction
Literature Review
Objectives
Proposed Model
Experimental Setup
Result Analysis
Conclusion & Future Scope
Reference
Acknowledgement

Dept. of CSE, University of Engineering &

Management Jaipur
 Introduction
• There are many different types of occupations in the world, but agriculture is the most common. The Indian
economy, which is heavily reliant on agriculture, is no exception. Global production exceeds 300 million
metric tons and is an important nutrition and calorie provider for humanity (Pareek 2016).
• The applications of image processing and artificial intelligence in agriculture in identifying and classifying
pests and diseases of plants and fruits have increased and research in this field is ongoing. In this paper, we
use the convolution neural network (CNN) methods
• Deep learning models perform better at mapping functions and thus generate better features.
• Using the Deep Learning approach, the proposed research methodology focuses on the classification and
identification of healthy and disease-infected leaf conditions. The architecture used in this study is VGG16
which is a VGG Network Group of the Convolutional Neural Network architecture models. Also different
types of CNN architectures like ResNet50 and GoogleNet are used in this work.
 Literature Review
• V. Suresh et al. in 2020 [1], they proposed a website-based image processing and machine learning

methodology for plant disease detection. The methodology involves gathering images, processing

them using a trained classifier through a website, and then redirecting to a page with appropriate

pesticides for treatment. This approach not only detects diseases but also suggests suitable

pesticides for treatment.

• Mrs. Shruthi U et al. in 2019 [2], they proposed machine learning methods for plant disease

classification, focusing on data acquisition, processing, segmentation, feature extraction, and image

classification. Their survey showed that Convolutional Neural Networks achieved high accuracy in

identifying multiple disease crops.

• Rajleen Kaur and Dr. Sandeep Singh Kang 2015[3] proposed for automatic disease detection and

disease part of the plant leaf images and also crop agriculture production. It's achieved with

computer advancement technology that allows agriculture to develop production.

Dept. of CSE, University of Engineering &

Management Jaipur
 Objectives
The general objectives of the study is to develop a reliable, convenient and accurate system. The study
has the following specific objectives:
• Identify the disease by visualizing the Image from the Given dataset based on the respective
category(Early Blight, Healthy, Late Blight).
• Characterization of the symptoms and damage caused by each type of potato disease, including
their impact on yield and economic outcomes.
• Development of a classification system based on the characteristics of potato diseases from the
images provided, and the methods of disease diagnosis.
• Provision of guidance to growers, researchers, and other stakeholders on effective disease
management and control strategies based on the classification system.

Dept. of CSE, University of Engineering &

Management Jaipur
 Proposed Model
• Image Dataset: This category would classify various images of potato leafs which contains healthy

and disease affected leafs

• Image Validation: This category would classify potato leafs on the basis of disease affected

and healthy leafs.

• Sequential Modelling: In this category the image which have been taken from images validation will

pass through certain defined conditions through the model and will give certain results.

• Evaluation Matrix: In this category the result which is obtained from sequential modelling will be

displayed in an array format.

Dept. of CSE, University of Engineering &

Management Jaipur
Experimental Set-up

Dept. of CSE, University of Engineering &

Management Jaipur
Result: Example

Dept. of CSE, University of Engineering &

Management Jaipur
 Result Analysis
 Prediction Probabilities: Late Blight (Class 2) has the highest probability of approximately

99.78%.Healthy (Class 1) has a relatively low probability of approximately 0.22%.Early Blight

(Class 0) has a very low probability of approximately 0.0036%.

 Interpretation: High probability for Late Blight indicates that the input is likely affected by this

disease. Low probabilities for Healthy and Early Blight suggest a low likelihood of these diseases.

 Implications: Accurate and early detection of Late Blight using the model can help farmers take

timely preventive measures. Late Blight is a severe disease that can cause crop damage and yield

loss, underscoring the importance of reliable classification tools.

Dept. of CSE, University of Engineering &

Management Jaipur
 Limitation
• Data limitations: Deep learning models for potato leaf disease diagnosis may suffer from overfitting,
reduced accuracy, and limited generalizability due to a lack of diverse and representative data.
• Interpretability: The lack of interpretability in deep learning models can create challenges in
understanding and validating the model's predictions, particularly in contexts where interpretability is
crucial.
• Complexity and computational resources: The computational complexity and resource requirements of
deep learning models, particularly CNNs, can pose challenges for deploying and running them in resource-
constrained environments like field-based settings or low-powered devices.
• Generalization to new diseases: Deep learning models trained on known diseases may not generalize to
newly emerging or rare diseases, necessitating retraining or fine-tuning with additional data for effective
classification.

Dept. of CSE, University of Engineering &

Management Jaipur
 Conclusions & Future Scope
• We have used the concept of CNN and have developed a model to classify conditions in the potato

leaves like early blight, late blight and healthy achieving classification accuracy of around 96.93%

• The proposed framework able to achieve an accuracy of 90% and above.

• Improved Accuracy: Deep Learning and CNN models can be refined with advanced architectures

to achieve higher accuracy in predicting potato leaf diseases.

• Real-time Prediction: Real-time prediction models using mobile devices or drones can quickly

detect and treat potato leaf diseases in the field.

Dept. of CSE, University of Engineering &

Management Jaipur
 References
• U. Shruthi, V. Nagaveni, and B. K. Raghavendra, A Review on Machine Learning
Classification Techniques for Plant Disease Detection 2019, 2019 5th Int. Conf. Adv.
Comput. Commun. Syst., 281–4.
• Suresh V, Gopinath D, Hemavarthini M, Jayanthan K, and Mohana Krishnan, Plant disease using
image processing 2020, IJERT, 9, 78-82.
• S. Kaur, S. Pandey, and S. Goel, Plants Disease Identification and Classification Through Leaf
Images: A Survey 2019, Arch. Comput. Methods Eng., 26, 507–30.
• H. Waghmare, R. Kokare, and Y. Dandawate, Detection and classification of diseases of Grape
plant using opposite color Local Binary Pattern feature and machine learning for automated
Decision Support System 2016, 3rd Int. Conf. Signal Process. Integ. Networks,
• R. Kaur and S. S. Kang, An enhancement in classifier support vector machine to improve plant
disease detection, Proc. 2015 IEEE 3rd Int. Conf. MOOCs, Innov. Technol. Educ.,135–40.

Dept. of CSE, University of Engineering &

Management Jaipur
 Acknowledgement

We would like to express my gratitude towards Prof. Hriday Banerjee who has been a precious help

in my voyage of finishing this undertaking. Without his assistance, direction, and support, making

this project a reality would have been unimaginable. His opportune criticism was tremendously

essential for me and helped me greatly to reach the targets set for the project tasks. Likewise, we

would also like to express gratitude toward Dr. Minal Kanti Sarkar for his commitment and

supporting this project as my second reader. In conclusion, we would like to thank the personnel and

staff of the Department of Computer Science Engineering at University of Engineering and

Management for their ceaseless regulatory help.

Dept. of CSE, University of Engineering &

Management Jaipur
Dept. of CSE, University of Engineering &
Management Jaipur