GSSS INSTITUTE OF ENGINEERING & TECHNOLOGY FOR WOMEN

(Affiliated to VTU, Belagavi, Approved by AICTE, New Delhi & Govt. of Karnataka)
(Accredited with Grade ‘A’ by NAAC)

DEPARTMENT OF INFORMATION SCIENCE AND ENGINEERING

(Accredited by NBA, New Delhi, Validity 01.07.2023 to 30.06.2026)

MAJOR PROJECT ON
“LEAF LENS : PLANT DISEASE DETECTION USING LEAF IMAGE
ANALYSIS”

Submitted by :- Under the Guidance of :-

Anushree K [4GW22IS011]
Lahari C.N [4GW22IS027] Mrs. Padmaja K
Pooja K.G [4GW22IS039] Assistant Professor
Thrishma S Singh [4GW22IS061] DEPT of ISE, GSSSIETW, Mysuru
 CONTENTS
Introduction
Motivation for the Project
Literature Survey
Problem Statement
Objectives
System Requirements:
Hardware Requirements
Software Requirements
Functional Requirements
Non-Functional Requirements
Methodology
Expected Outcomes
References
 INTRODUCTION
 Plant diseases are one of the leading causes of crop loss in
worldwide, significantly impacting food production, agricultural
sustainability, and the economy.
 Plant disease detection using leaf image analysis combines
computer vision and machine learning for advanced agricultural
diagnostics.
 The LEAF LENS project focuses on analyzing leaf images to
identify diseases early, enabling farmers to take quick action
 Traditional methods of disease detection are slow and require expert
knowledge, leading to delayed responses and crop loss.
 It reduces the use of harmful pesticides and promotes eco-friendly
farming.
 It works faster and more accurately than checking plant diseases
manually.
 Motivation for the Project
 Preventing Crop Loss: Plant diseases can cause huge losses in
agricultural yield, so early detection helps save crops.

 Simplifying Disease Identification: Traditional methods need experts

and take time, but this technology makes detection faster and easier.

 Empowering Farmers: Farmers get quick, accurate information about

plant health, enabling timely action.

 Reducing Pesticide Use: Identifying diseases precisely helps avoid

excessive use of chemical pesticides, protecting the environment.

 Food Security: Ensure better yields to meet the growing global demand
for food.
 Literature Survey
TITLE AUTHOR METHODOLOGY REFERENCE YEAR LIMITATIONS
Plant Leaf Disease Prediction Meroua Belmir, Utilized a Convolutional Neural IEEE Xplore The study relies on a
and Classification Using Deep Wafa Difallah,
2023
Network (CNN) model trained on limited dataset with
Learning Abdelkader the Plant Village dataset to images collected under
Ghazli. classify plant diseases with high controlled conditions,
accuracy which may not fully
generalize to real-world
agricultural environments.

Nitin Lokhande, Compared CNN models IEEE Xplore 2024 The study relies on a
Comparative Analysis of
Vijaya Thool, (AlexNet and ResNet-50) for limited dataset with
Different Plant Leaf Disease Pratap Vikhe. classifying maize and soybean images collected under
leaf diseases using transfer controlled conditions,
Classification and Detection which may not fully
learning and image preprocessing generalize to real-world
Using CNN techniques. agricultural environments.
Deep Learning-Based BeddaHana, Evaluated multiple deep learning Journal of 2024 The dataset is relatively small,
Detection of Apple Leaf models (CNN, VGG16, ResNet50, Medical leading to possible overfitting, and
Saadna Yassmina. ResNet101) on the 'Plant Pathology Systems
Diseases Using Image real-world conditions such as
2020 - FGVC7' dataset to classify
Analysis lighting and environmental
apple leaf diseases, with DenseNet
achieving the best validation variations were not fully considered.
accuracy of 97.18%.

D. Yaswanth, Dr. Deepak The model struggles to

Plant Leaf Disease
Chowdary,
Used transfer learning with pre- IEEE Xplore 2024 distinguish between diseases with
Detection Using S. Sai Manoj, Srikanth trained CNN models similar visual features, requiring
Learning Yadav M.
more diverse datasets and
Transfer (ResNet50) on the Plant Village
additional fine-tuning for better
Approach dataset for plant leaf disease accuracy in real-world
agricultural conditions.
classification, achieving a
validation accuracy of 91% and
a testing accuracy of 89%.
 M. Dhilsath Fathima,
Transfer Learning for
Akash
Utilized a hybrid deep IEEE Xplore 2023 The model relies on a
Multi-Crop Leaf Gupta, Kartik Jain. learning model controlled dataset, which
Disease Prediction combining ResNet and
Using ResNet and ConvNet with transfer may limit its effectiveness in
ConvNet learning on the Plant real-world agricultural
Village dataset, environments, and additional
achieving an accuracy
fine-tuning may be required
of 98.2% for multi-crop
leaf disease for better generalization
classification. across diverse
crop conditions.

Aquatic Plant R. Aishwarya, Yogitha Applied deep learning Springer 2023 The model's
Disease Detection R., Froila Stephanie P.A,
Chegoni Dhanusha models, including CNN, effectiveness is limited
Using Deep Learning
Srinivas SVM, and KNN, by dataset constraints,
combined with K-means and its sensitivity to
clustering and PCA for noise may reduce
feature extraction to stability, requiring
detect diseases in further refinement for
aquatic plants. real-world deployment
 Utilized deep learning 2024 The model is
Leaf Lens: An P. Yogendra Prasad, IEEE Xplore computationally
techniques, specifically the
Rayachoty Chandana VGG19 model, for plant intensive, requiring
Intelligent Vision for significant resources for
disease detection based on leaf
Plant Disease Diagnosis Priya, M. Ramu, real-time application,
images, achieving high
and may need fine-
Using Deep Learning Pasupuleti Anusha, accuracy through automated tuning for different plant
feature extraction and species and real-world
Sanjamala Reshma, B. classification. agricultural conditions.
Likhith Reddy.

Plant Disease Detection 2024 The model requires a

Vela Chelamala Ganesh Implemented a deep IEEE Xplore high-quality dataset for
and Pesticide
Recommendation Using Reddy, Arul Xavier V M, learning-based system optimal performance,
Deep Learning and real-world
Shyni S S. using a modified
application may be
EfficientNetB5 model affected by
environmental variations
for plant disease
and computational
detection and pesticide constraints.
recommendation,
achieving over 96%
accuracy.
Plant Disease A. Lakshmanarao, Used ConvNets for ACM 2021 The model's
Prediction and M.Raja Babu, plant disease performance is
Classification Using T. Srinivas Ravi Kiran. classification, dependent on the
quality and diversity of
Deep Learning leveraging deep the dataset, and its
ConvNets. learning to improve computational
accuracy and requirements may
efficiency in disease hinder real-time
prediction. deployment in
agricultural settings.

Automatic Prediction of Asta Lakshmi M, Dr. V. Conducted a comprehensive ACM 2021 The study lacks
Plant Leaf Diseases Gomathi. review of deep learning-based experimental
Using Deep Learning plant disease detection models, implementation and
Models: A Review highlighting the effectiveness comparative real-world
of transfer learning with testing, focusing mainly
DenseNet and the use of on reviewing existing
NVIDIA Jetson Nano for real- approaches rather than
time implementation. proposing a novel
model.
Plant-Leaf Disease Varun Tiwari, - Utilizes IEEE Xplore 2024 Limited training data
Detection: A Convolutional Neural for rare diseases.
Comparative Analysis
Amit Kumar Networks (CNNs) for - Generalization issues
with Machine Bairwa, feature extraction and across different crops
Learning and Deep Shashi Bhushan classification. and environmental
Learning Methods Sharma - Evaluates traditional conditions.
ML models like - High computational
Decision Trees and requirements for deep
Support Vector learning models.
Machines (SVMs).
- Uses publicly
available datasets of
diseased plant leaves.

Deep Learning-Based Kelothu Shivaprasad, - Uses deep learning- IEEE Xplore - Challenges in dataset
Plant Leaf Disease Ankita Wadhawan. based approaches, 2023 quality and annotation.
mainly CNN
Detection architectures. - High computational costs
- Image preprocessing for real-time
techniques applied to implementation.
improve classification
- Difficulty in model
accuracy.
- Performance interpretability for end-
comparison with users.
traditional ML
approaches.
Detection of Plant Disease and Pratibha Nayar, - Focuses on dl models for
IEEE Xplore 2022 - Computational complexity
due to deep model architecture.
Pests Using Coherent Deep Shivank Chhibber, detecting both plant diseases
Learning Algorithms Ashwani Kumar Dubey. and pests.
- Difficulty in differentiating
- Uses image classification and visually similar diseases.
object detection techniques. - Requirement for large,
- Employs CNNs for improved diverse datasets for model
accuracy. generalization.
Plant Disease Detection and Vela Chelamala Ganesh -Integrates dl-based plant - Challenges in real-time
Pesticide Recommendation disease classification with
IEEE Xplore 2024 deployment due to
Using Deep Learning
Reddy,
recommendation system for computational needs.
Arul Xavier V M ,
pesticides. - Possible bias in pesticide
Shyni S S. recommendations.
- Uses CNN-based feature
- Limited scalability for new
extraction and a decision- disease variants.
support model for pesticide
selection.

Plant Leaf Disease Detection D. Yaswanth, - Utilizes transfer learning

Using Transfer Learning S. Sai Manoj with pre-trained deep
IEEE Xplore 2024 - Dependency on pre-trained
models may lead to suboptimal
Approach Srikanth, learning models such as
Yadav. M ResNet or VGG for plant performance on unseen plant
disease classification. diseases.
- Data augmentation
- Need for high-quality
techniques used to enhance
training efficiency. annotated datasets.
- Issues with model
explainability for end-users.
 PROBLEM STATEMENT
 To address these challenges, there is an urgent need for an automated, accurate, and eco-
friendly plant disease prediction system. This project aims to leverage deep learning
techniques to develop a smart system capable of analyzing leaf images to detect diseases
efficiently. By providing farmers with a fast and reliable disease diagnosis tool, this system
will help reduce reliance on harmful pesticides, improve agricultural productivity, and
promote sustainable farming practices.

 Plant diseases are a major threat to global agriculture, causing substantial crop losses and
economic damage. Traditional disease detection methods rely on visual inspection by
farmers and agricultural experts, which is often subjective, time-consuming, and prone to
human error. This delay in diagnosis can lead to the rapid spread of infections, further
reducing crop yields.

 Additionally, conventional disease management heavily depends on chemical pesticides,

which pose serious health risks to farmers, including respiratory problems, skin irritation,
and even cancer. Excessive pesticide use also leads to environmental degradation by
contaminating soil and water and disrupting the natural balance of beneficial
microorganisms.
 OBJECTIVES
 To collect and analyze the dataset: Collect the dataset of leaf images
representing various plant diseases and analyze the dataset to identify
patterns and features.
 Testing and training the model: train and test a deep learning-based model
using the collected dataset to predict plant diseases from leaf images.
 Design solution: to design user friendly interface that allows farmers to
easily capture and upload images of leaf for disease detection.
 Deployment: Deploy the trained model that can be easily adopted by farmers
with limited financial resources, promoting sustainable agriculture.
 SYSTEM REQUIREMENTS
HARDWARE REQUIREMENTS :

 Standard computer or server with internet access.

 CPU with multi-core support for deep learning

computations

 GPU for accelerating deep learning tasks.

 High-capacity storage (for large datasets).

 RAM (8GB or more recommended).

 Network connectivity for cloud-based operations.

 Reliable power supply.

 Portable device for field data collection.

 Backup storage system for data safety.

 SOFTWARE REQUIREMENTS

 Python programming language.

 Keras and TensorFlow libraries for machine

learning.

 Web application development library (streamlit)

 Image processing libraries (OpenCV, PILLOW).

 Dataset analysis libraries (Pandas, NumPy).

 Cloud storage (Google Drive)

 Dataset and microscopy images (Kaggle).

 Google collab for model training.

 FUNCTIONAL REQUIREMENTS
 Leaf Image Input: The system must allow users to upload leaf image for analysis and disease
prediction.

 Disease Prediction: The system must analyze the uploaded images and predict the disease present in
the uploaded image

 Disease Classification: The system must classify predicted disease into a specific category (eg:
bacterial, fungal)

 Fast Diagnosis: The system should provide diagnostic results within a few seconds after the image is
submitted for analysis.

 Model Training: The system must allow for the continuous training and improvement of the machine
learning model using a large dataset.
 NON-FUNCTIONAL REQUIREMENTS

 Accuracy and Performance: The system must maintain high accuracy, and system should be

able respond quickly after uploading image.

 Usability: The interface should be user-friendly, and allows user to easily upload images and

view results.
 Response Time: The system should deliver predicted results within seconds after an image is

submitted, ensuring quick feedback for decisions.

 Reliability: The system should operate continuously without failures, and should be highly

available
 Maintainability: System easy to maintain and repair.
 Portability: The system should be deployable across various platforms, including cloud-based

environments, desktop computers and etc,

 Scalability: System scalable for larger installations.
 METHODOLOGY
Start

Image pre -processing and labelling

Input image

Augmentation process

Neutral Network Training

Testing

Classified Disease Healthy image

Output result

End
Dataset Description:-
Sl.no Plant name Heathy or diseased Disease Name
1 Apple Diseased Apple scab
2 Apple Diseased Black rot
3 Apple Diseased Cedar apple rust
4 Apple Healthy -
5 Blueberry Healthy -
6 Cherry Healthy -
7 Cherry Diseased Powdery mildew
8 Grape Diseased Black rot
9 Grape Diseased Esca Black measles
10 Grape Healthy -
11 Grape Diseased Leaf blight Isariopsis leaf spot

12 Orange Diseased Haunglongbing

13 Peach Diseased Bacterial spot
14 Peach Healthy -
15 Pepper Diseased Bell bacterial spot
16 Pepper Healthy -
17 Potato Diseased Early blight
18 Raspberry Healthy -
19 Strawberry Leaf scorch Leaf scorch
20 Squash Diseased Powdery mildew
1. Image Preprocessing and labelling
The raw images collected from the dataset might contain noises and it is essential to preprocess them before fitting them into
the learning module.

2. Input image
The proposal starts with collecting the input images representing different types of leaves like apple, orange, grape, peach. These raw images
can be collected from Kaggle dataset.
3. Augmentation process

Data augmentation is a technique for increasing the number of images in a database. Various operations such as shifting, rotating, zooming, and
flipping are applied to image datasets to diversify our dataset.

4. Neural network training and Testing

 From an ideal dataset, 80% of photos are taken for training and 20%
for testing.
 Validation data is used to check accuracy by applying the predict function
and accurately extracting features.
 Images are taken for confirming detection once validation provides
good results.
 Finally, characteristics are retrieved to determine whether

or leaves are infected or not.

5. Deep Learning models :

• Why Vgg16 and Vgg19 ? (Visual Geometry Group)

VGG16 and VGG19 are well-suited for image classification tasks, including plant disease detection, because they can learn hierarchical features from images,
allowing them to identify subtle patterns and differences between healthy and diseased leaves.
While both models are similar, VGG19 has more layers (19 compared to 16), which can lead to slightly improved accuracy, but also requires more computational
resources and training time.
The convolutional layers in VGG16 and VGG19 learn to extract relevant features from images, such as edges, textures, and shapes, which are then used to
classify the images.

• Why ResNet50 ?
ResNet50 is used for plant disease detection because its architecture, particularly the use of residual connections, allows for training deeper networks, leading to
improved accuracy and the ability to learn complex features from plant leaf images, even when diseases have similar visual characteristics
One study achieved 98.98% accuracy using ResNet50 for crop disease prediction.
VGG achieved good accuracy i.e. 99%
 EXPECTED OUTCOMES:

1. Automated Disease Detection:

o Uses leaf images to detect plant diseases accurately and instantly.
o Eliminates the need for slow and sometimes inaccurate traditional methods.

2. Early Intervention & Crop Protection:

o Identifies diseases in early stages to prevent their spread.
o Reduces crop losses and improves overall yield.

3. Reduced Pesticide Use & Environmental Benefits:

o Prevents excessive pesticide use, reducing health risks for farmers.
o Minimizes soil and water contamination, preserving farmland fertility.

4. User-Friendly Design:
o Simple interface for easy use by farmers with little technical knowledge.
o Works with smartphone cameras for instant disease diagnosis.

5. Versatile & Scalable Model:

o Deep learning model trained on multiple plant species.
o Improves over time with more data for better accuracy.
 REFERENCES:

1] M. Belmir, W. Difallah, and A. Ghazli, "Plant Leaf Disease Prediction and Classification Using Deep Learning," 2023
International Conference on Artificial Intelligence and Machine Vision (AIMV), IEEE, 2023.

[2] N. Lokhande, V. Thool, and P. Vikhe, "Comparative Analysis of Different Plant Leaf Disease Classification and
Detection Using CNN," International Journal of Computer Science and Engineering, vol. XX, no. X, pp. XX-XX, 2024.

[3] B. Hana and S. Yassmina, "Deep Learning-Based Detection of Apple Leaf Diseases Using Image Analysis," 2024
International Conference on Computer Vision and Pattern Recognition, IEEE, 2024.

[4] D. Yaswanth, E. D. Chowdary, S. S. Manoj, and S. Yadav M, "Plant Leaf Disease Detection Using Transfer Learning
Approach," International Journal of Machine Learning and Cybernetics, vol. XX, no. X, pp. XX-XX, 2024.

[5] M. D. Fathima, A. Gupta, and K. Jain, "Deep - Transfer Learning for Multi-Crop Leaf Disease Prediction Using ResNet
and ConvNet," International Journal of Artificial Intelligence and Data Science, vol. XX, no. X, pp. XX-XX, 2024.

[6] R. Aishwarya, Y. R., F. S. P. A, and C. D. S, "Aquatic Plant Disease Detection Using Deep Learning," 2023 International
Conference on Emerging Trends in Artificial Intelligence and IoT, IEEE, 2023.
[7] P. Y. Prasad et al., "Leaf Lens: An Intelligent Vision for Plant Disease Diagnosis Using Deep Learning," 2024 International Conference on Advances in Machine
Learning and Data Science, IEEE, 2024.

[8] V. C. Ganesh Reddy, A. Xavier V. M, and S. S. Shyni, "Plant Disease Detection and Pesticide Recommendation Using Deep Learning," International Journal of
Agricultural Informatics and Technology, vol. XX, no. X, pp. XX-XX, 2024.

[9] A. Lakshmanarao, M. Raja Babu, T. Srinivas Ravi Kiran, "Plant Disease Prediction and Classification Using Deep Learning ConvNets," 2021 International
Conference on Artificial Intelligence and Machine Vision (AIMV), IEEE, 2021.

[10] A. L. M and V. Gomathi, "Automatic Prediction of Plant Leaf Diseases Using Deep Learning Models: A Review," 2021 5th International Conference on
Electrical, Electronics, Communication, Computer Technologies and Optimization Techniques (ICEECCOT), IEEE, 2021.

[11] V. Tiwari, A. K. Bairwa, and S. B. Sharma, "Plant-Leaf Disease Detection: A Comparative Analysis with Machine Learning and Deep Learning Methods," in
2024 International Conference on Modeling, Simulation & Intelligent Computing (MoSICom), Dubai, UAE, Dec. 9-11, 2024, pp. 166–170.

[12] K. Shivaprasad, "Deep Learning-based Plant Leaf Disease Detection," in 2023 7th International Conference on Intelligent Computing and Control Systems
(ICICCS), 2023, pp. 361–365.

[13] P. Nayar, S. Chhibber, and A. K. Dubey, "Detection of Plant Disease and Pests using Coherent Deep Learning Algorithms," in 2022 14th International
Conference on Computational Intelligence and Communication Networks (CICN), Al-Khobar, Saudi Arabia, 2022, pp. 156–160.

[14] V. C. G. Reddy, A. X. V. M, and S. S. S, "Plant Disease Detection and Pesticide Recommendation Using Deep Learning," in 2024 10th International Conference
on Advanced Computing and Communication Systems (ICACCS), Coimbatore, India, 2024, pp. 2543–2547.

[15] D. Yaswanth, S. S. Manoj, S. Y. M, and E. D. Chowdary, "Plant Leaf Disease Detection Using Transfer Learning Approach," in 2024 IEEE International
Students’ Conference on Electrical, Electronics and Computer Science (SCEECS), [Location if specified], 2024, pp.