Scribd
Upload
27 views11 pages

Plant Disease Detection

The document discusses using machine learning for disease detection in plant leaves. It details challenges with manual detection, the problem statement, training a CNN model on a cotton leaf disease dataset, achieving 95.25% accuracy, and future work including improving specificity, mobile implementation, and integrating other sensors.

Uploaded by

curiousastrophile001
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
27 views11 pages

Plant Disease Detection

The document discusses using machine learning for disease detection in plant leaves. It details challenges with manual detection, the problem statement, training a CNN model on a cotton leaf disease dataset, achieving 95.25% accuracy, and future work including improving specificity, mobile implementation, and integrating other sensors.

Uploaded by

curiousastrophile001
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

DI SEASE

DETECTION
IN LEAF USING ML
Team Members :
Navjeet Lohan ( 21104062 )
Navjot Singh ( 21104063 )
Yash Goyal ( 21104114 )
Shivanshu Kr ( 21112079 )

Supervised By:
Dr. Deepti Kakkar
Dr. B.S. Saini
(Department of ECE)
TABLE OF CONTENT
Challenges of Manual Disease Detection
Problem Statement
Training Dataset
Model used
Results
Future Scope
References
CHALLENGES OF MANUAL DISEASE
DETECTION
Unfortunately, crops are susceptible to various

Vision
diseases. Traditionally, detecting these diseases
has been a manual process, often leading to:

Inaccuracy: Early signs of disease can be subtle,


leading to missed diagnoses by human inspectors.
Inefficiency: Extensive fields require significant
time and manpower for manual inspection.
Subjectivity: Diagnosis can be subjective,
depending on the inspector's experience and
training.
PROBLEM STATEMENT

1 To choose an efficient and flexible


technique for image classification and
disease detection.

2 Constructing a model for training dataset and


input image feature extraction.

3 Finally, ensuring that model is reliable for


feature extraction and can achieve better
accuracy
TRAINING DATASET

Taken cotton leaf plant for disease


detection.
The Dataset is collected from open
source website “Kaggle”.
https://www.kaggle.com/datasets/janmejaybhoi/cotton-
disease-dataset

The dataset is splited in 3 types:


1. Train data
2. Test data
3. Validation data

The dataset contains around 2k images


and has size of 156 MB
Input (128x128x3)

CNN MODEL |
Conv2D (64 filters, 3x3)
|
Conv2D (64 filters, 3x3)

Architecture of our CNN model: |


MaxPooling2D (2x2)
|
Conv2D (128 filters, 3x3)
Input layer: Image dimensions of 128x128 pixels. |
Conv2D (128 filters, 3x3)
Multiple convolutional layers with increasing filters to extract |
MaxPooling2D (2x2)
features. |
Max pooling layers to reduce spatial dimensions and control Conv2D (128 filters, 3x3)
|
overfitting. Conv2D (128 filters, 3x3)
|
Dropout layers to prevent overfitting by randomly dropping neurons MaxPooling2D (2x2)
|
during training. Conv2D (128 filters, 3x3)
Dense layers for classification. |
Conv2D (128 filters, 3x3)
Used ReLU activation functions, which introduce non-linearity into |
MaxPooling2D (2x2)
the model. |
Conv2D (128 filters, 3x3)
|
Conv2D (128 filters, 3x3)
|
MaxPooling2D (2x2)
|
Dropout (0.5)
|
Flatten
|
Dense (64 units)
|
Dropout (0.5)
|
Dense (4 units)
PERFORMANCE AND RESULTS

Created CNN model and achieved accuracy of 95.25%.

Input image can be tested for disease on jupyter platform.

Disease detection in Leaf github Repository


Densenet121 Model
Modification in densenet121 model.
Added average pooling layer and
batch normalization layer and
changed filters in some convolution
layers.
Results in 97% accuracy
FUTURE SCOPE
Improved Disease Specificity: Currently, the model might classify diseases into broad
categories. Future work could involve collecting more data and fine-tuning the model to
differentiate between specific diseases with greater accuracy.

Mobile Platform Implementation: Optimizing the model for deployment on mobile


devices would make the disease detection app more accessible to farmers, especially in
resource-limited areas. This might involve techniques like model compression or
quantization to reduce computational requirements.

Integration with other Sensors: The model could be integrated with additional sensors
that measure factors like soil moisture or nutrient levels. This combined data can
provide a more comprehensive picture of plant health and allow for the development of
more sophisticated disease prediction models.
REFERENCES:

https://ieeexplore.ieee.org/document/9777219

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8659659/

https://www.kaggle.com/datasets/singhakash/cotton-disease-dataset

https://www.researchgate.net/publication/362742399_Convolutional_Neural
_Network_CNN_A_comprehensive_overview
THANK YOU

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy