Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 29

CSE499B

AGRICULTURE CROP DISEASE DETECTION USING

IMAGE PROCESSING
OUR TEAM

Md Saleh Ahmed , Id:1821259042


Sadia Mahjabin Sanchary , Id:1912138042
Susmita Roy Happy , Id:1911678043
Introduction
Agriculture is the backbone of our global food
supply chain. It sustains not only our livelihoods
but also the foundation of food security. With
the world's population steadily growing, the
importance of agriculture has never been
greater. Ensuring a stable and bountiful crop
yield is a critical aspect of this equation.
The rising challenge of crop diseases

However, agriculture faces numerous challenges, and one of


the most pressing is the increasing prevalence of crop
diseases. These diseases threaten not only crop yields but also
the livelihoods of millions of farmers worldwide. Timely
detection and effective management of crop diseases are
pivotal in ensuring food security for our growing population.

This project endeavors to tackle this challenge head-on, using


cutting-edge image processing techniques to revolutionize the
way we detect and respond to crop diseases.
Problem Statement
Crop diseases can lead to significant crop yield loss:
Crop diseases, caused by various pathogens and environmental factors, have a
detrimental impact on agricultural productivity.
They manifest in the form of leaf spots, wilting, discoloration, and deformities,
ultimately resulting in yield reduction and financial losses for farmers.

Emphasize the need for timely disease detection:


Timely detection of crop diseases is crucial to prevent or mitigate yield loss.
Early identification enables prompt intervention, reducing the spread of diseases
and the need for extensive crop treatment.
Current detection methods are often labor-intensive, time-consuming, and reliant
on visual inspection, making them inefficient and prone to human error.

By addressing this problem, we aim to revolutionize disease detection in


agriculture and empower farmers with efficient and accurate tools to safeguard
their crops and livelihoods.
Importance of Solving This Problem
The impact of crop diseases on food production:
Crop diseases have a profound and detrimental impact on food
production.
They directly reduce crop yields, affecting the quantity and quality of
food available.
As a result, crop diseases contribute to food scarcity and increased prices
in the global market.
How early detection can mitigate these issues:
Early detection of crop diseases is akin to an early warning system for
the agricultural industry.
It enables farmers to take proactive measures to manage and contain the
spread of diseases.
This can include targeted pesticide use, isolated crop removal, or
alternative planting strategies, ultimately reducing the impact on yield.
The potential benefits for farmers and the
agricultural industry:

The adoption of efficient disease detection methods can


significantly increase crop yields and quality.
Reduced yield loss leads to improved economic stability for
farmers and a more sustainable agricultural industry.
Increased food production also plays a critical role in addressing
global food security challenges.

Solving the challenge of timely crop disease detection has far-


reaching implications, from addressing the immediate concerns of
individual farmers to contributing to the broader goal of ensuring
food security for our growing population.
Work Flow
An overview of the project's two parts:
This project is structured into two key phases, each with its specific
objectives and tasks.

499A: Planning, data collection, and decision-making:


Part one encompasses the initial stages of our project.
It is primarily focused on meticulous planning, data collection, and strategic
decision-making.
During this phase, we lay the foundation for the implementation phase.

499B: Implementation, model development, and


completion:
Part two marks the core of our project's technical execution.
It includes the implementation of our disease detection system, model
development, and project completion.
Here, we will put our plans into action, develop and fine-tune our
detection models, and prepare for the project's conclusion.
Dataset
We utilized the PlantVillage Dataset, available on Kaggle, for our project on plant disease
prediction. The dataset comprises a vast collection of images depicting various plant
diseases and healthy plant samples.

Number of Images:
• The PlantVillage Dataset contains over 50,000 images, making it a rich resource for
training and evaluating machine learning models.
Classes of Diseases Covered:
• The dataset covers a wide range of plant diseases across multiple plant species,
including but not limited to:
• Tomato: Early Blight, Late Blight, Bacterial Spot, Leaf Mold, Septoria Leaf Spot, etc.
• Potato: Early Blight, Late Blight
• Apple: Apple Scab, Cedar Apple Rust, Healthy, etc.
• Grape: Black Rot, Esca, Healthy, etc.
• And many more.
Preprocessing Steps Applied:

Prior to training our machine learning model, we performed several preprocessing steps on the dataset,
including:
• Image resizing to a standardized resolution to ensure uniformity.
• Normalization of pixel values to the range [0, 1] to facilitate convergence during model training.
• Splitting the dataset into training, validation, and test sets to assess the model's performance on unseen
data.
WORK FOR 499B
Image Data Preprocessing:
Part Two marks a crucial phase in the development of our disease detection system, where we transition from planning and data
collection to the technical execution of our project. This phase encompasses the following key tasks:
Data preprocessing, resizing, Noise Reduction and Normalization:
We will delve into the meticulous preparation of our dataset.
Data preprocessing is a foundational step that involves activities such as image resizing to a consistent format and normalization
to ensure uniformity in data distribution. These measures enhance the quality and suitability of our dataset for training and
evaluation.
Model exploration and optimization:
Model exploration entails fine-tuning and optimization of our models to adapt them to the specific demands of our crop disease
detection task. We will explore various techniques to improve model accuracy and efficiency .
Model Architecture

Description of the CNN Architecture Used:


Our CNN architecture is meticulously crafted to
address the unique challenges of plant disease classification. It encompasses multiple layers
meticulously designed to extract intricate features from input images and make precise
predictions across various disease categories. Below, we provide an exhaustive breakdown of
this architecture

Layers:

1. Convolutional Layers
2. Pooling Layers
3. Dropout Layers
4. Fully Connected Layers
Activation Functions:
• The activation function used in the convolutional layers is the Rectified Linear Unit (ReLU),
which introduces non-linearity and helps the network learn complex patterns from the input
data.
• The final dense layer utilizes the softmax activation function, which converts the raw output
scores into probabilities, indicating the likelihood of each class.

Brief Explanation of How the Model Learns to Classify Diseases:

• The CNN architecture leverages the hierarchical structure of the layers to automatically
learn discriminative features from input images.
• Initially, low-level features such as edges and textures are detected by the early
convolutional layers.
• As the network progresses through subsequent layers, higher-level features relevant to
disease classification, such as lesion patterns or discolorations, are extracted.
• The fully connected layers at the end of the network utilize these learned features to make
predictions, mapping them to the appropriate disease classes based on the training data.
Loading The Dataset:

The images cover 14 species of crops, including: apple, blueberry, cherry, grape,
orange, peach, pepper, potato, raspberry, soy, squash, strawberry and tomato. It
contains images of 17 basic diseases, 4 bacterial diseases, 2 diseases caused
by mold (oomycete), 2 viral diseases and 1 disease caused by a mite. 12 crop
species also have healthy leaf images that are not visibly affected by disease.
Data Preprocessing:
Data preprocessing is a crucial step in preparing the raw input data for training the Convolutional Neural Network (CNN)
model. This process involves several steps aimed at enhancing the quality and usability of the dataset for effective
model training. Standardizing the size of input images to a uniform resolution (e.g., 224x224 pixels) ensures
consistency across the dataset.
Build the model:

Our Convolutional Neural Network (CNN) model is meticulously designed to effectively classify plant diseases from
input images. The architecture comprises multiple layers, each meticulously crafted to extract hierarchical features and
enable accurate classification.
Model Training:
Model Evaluation:
Model Accuracy:
Final Testing:
Web App:

In addition to developing a Convolutional Neural Network (CNN) model for plant disease prediction, we have also
created a user-friendly web application using Streamlit. The web app serves as an intuitive interface for users to
interact with the model, enabling them to upload images of plant leaves and receive instant predictions regarding
potential diseases.

Key Features:
Upload Image:
• Users can easily upload images of plant leaves directly from their devices using the web app's user interface.
• Prediction Display:
• Upon uploading an image, the web app utilizes the trained CNN model to predict the likelihood of various plant
diseases affecting the uploaded leaf.
• Visualization:
• The web app provides visual feedback, including the uploaded image and the corresponding disease prediction,
enhancing user understanding and engagement.
• User-Friendly Interface:
• Streamlit's intuitive design allows for the creation of a user-friendly interface with minimal code, enabling seamless
navigation and interaction for users of all levels.
Transfer Learning: Using VGG 16
Transfer learning using VGG16 involves leveraging a pre-trained VGG16 model, which has been
trained on a large dataset (usually ImageNet) for image classification tasks. Instead of training a neural
network from scratch, transfer learning allows us to use the knowledge (learned features) gained by
the VGG16 model on a different but related task.
Transfer Learning: Using ResNet-50
Transfer learning using ResNet-50 involves leveraging a pre-trained ResNet-50 model, which has been trained on
a large dataset (typically ImageNet) for image classification tasks.

Transfer Learning: Using MobileNet3


Transfer learning with MobileNetV3 is a powerful technique for leveraging pre-trained models to solve new
tasks with limited data. MobileNetV3 is a convolutional neural network architecture optimized for mobile and
embedded vision applications, known for its efficiency and accuracy
Image Segmentation: Image segmentation, a crucial computer vision task, finds
applications across medical imaging, autonomous driving, object detection, and image
editing.
Final Output:
Future Study
Potential areas for future research and improvement:
As our project advances, it is essential to identify areas where further research
and enhancement can be pursued. Several avenues for future study include:
Advanced Image Techniques: Exploring more advanced imaging
technologies, such as multispectral and hyperspectral imaging, to provide a
more comprehensive view of crop health and disease.
Data Augmentation: Experimenting with data augmentation techniques to
further diversify our dataset and improve model generalization.
Disease Localization: Developing models not only for disease detection
but also for pinpointing the exact location of diseases on plants.
Disease Forecasting: Moving beyond detection to predicting disease
outbreaks, enabling proactive disease management.
The inclusion of multispectral imaging for more
comprehensive disease detection:
In particular, we propose the incorporation of multispectral imaging to achieve
more comprehensive disease detection. Multispectral data captures information
beyond the visible spectrum, revealing subtler cues that could enhance our
disease detection accuracy.
These future research directions aim to advance our project's capabilities and
contribute to the continuous evolution of crop disease detection in agriculture.
Conclusion:
The key takeaways from your project:
The current work uses MATLAB image processing algorithms to detect different
types of disease. It entails loading the image, segmenting it, extracting its features,
and classifying it. Creating automatic detection systems that use cutting-edge
technologies, such as image processing, helps farmers identify illnesses early on
and provide valuable information for managing them. We should focus even more
on disease detection in our efforts.
The importance of early crop disease detection:
In conclusion, we emphasize the critical role of early crop disease detection in
ensuring global food security and the well-being of farmers.
Our project not only addresses a significant challenge but also offers a practical
solution that empowers farmers and advances agricultural practices.
References:
1. J. Smith, et al., "A Survey of Machine Learning Applications in Crop Disease
Detection," Journal of Agricultural Technology, vol. 45, no. 3, pp. 213-230,
2021.
2. A. Brown and C. Lee, "Transfer Learning with Convolutional Neural Networks
for Improved Crop Disease Detection," in International Conference on
Agricultural Engineering, 2020, pp. 135-149.
3. S. Patel, et al., "Image Preprocessing and Feature Extraction for Efficient Crop
Disease Detection," Journal of Agricultural Sciences, vol. 28, no. 4, pp. 367-
382, 2019.
4. R. Kumar and P. Sharma, "Multispectral Imaging for Enhanced Crop Disease
Diagnosis: A Review," Remote Sensing Applications, vol. 12, no. 2, pp. 153-
169, 2018.
5. National Agricultural Research Institute, "Annual Report on Crop Diseases and
Pests," NARI Publications, 2022.
6. S. S. Chawathe, "Rice Disease Detection by Image Analysis," 2020 10th
Annual Computing and Communication Workshop and Conference (CCWC),
Las Vegas, NV, USA, 2020, pp. 0524-0530, doi:
10.1109/CCWC47524.2020.9031140.
7. K. Ahmed, T. R. Shahidi, S. M. Irfanul Alam and S. Momen, "Rice Leaf Disease
Detection Using Machine Learning Techniques," 2019 International Conference
on Sustainable Technologies for Industry 4.0 (STI), Dhaka, Bangladesh, 2019,
pp. 1-5, doi: 10.1109/STI47673.2019.9068096.
Thank You!

You might also like