Leaf Pathogen identification Using Deep

Learning.
 S. Sharmila Dr. V. Jeyalakshmi
College of Engineering, Department of Electronics and Communications
Anna University Engineering, College of Engineering, Anna
Chennai- 600025, India University,Chennai- 600025, India
Email Id: [email protected] Email Id: [email protected]
 Abstract-- Leaf disease identification plays a crucial role in problems, including low productivity and financial losses to
agricultural management, enabling early detection and ranchers and farmers, are being caused by illnesses in
intervention to mitigate crop losses. In recent years, harvests.
advances in machine learning, particularly Convolutional
Neural Networks (CNNs), have shown promise in Leaf disease identification is a crucial aspect of
automating the process of leaf disease identification. We plant health management and agricultural sustainability.
propose a CNN-based approach for leaf disease
It involves the identification and classification of
identification, leveraging the power of deep learning to
accurately classify leaf images into healthy or diseased various diseases that affect the leaves of plants, such as
categories. Our model is trained on a diverse dataset fungi, bacteria, viruses, and nutrient deficiencies. Early
containing images of plant leaves such as tomato, blueberry, detection and accurate diagnosis of these diseases are
orange, peach, corn (maize), potato, raspberry, soy bean, essential for timely intervention and effective control
strawberry, and blueberry which is affected by various measures. Various methods, including visual
diseases such as scab, black spot, cedar spot as well as inspection, symptomatology, and advanced diagnostic
healthy leaves, enabling it to learn discriminative features tools like molecular techniques and machine learning
for disease detection. We design and implement a CNN algorithms, are employed to identify leaf diseases.
architecture tailored to the specific requirements of leaf
disease identification, incorporating techniques such as data
Proper identification allows farmers and gardeners to
augmentation, transfer learning, and fine-tuning to improve implement appropriate treatment strategies, including
model performance and generalization ability. The chemical treatments, cultural practices, and biological
experimental results demonstrate the efficacy of our CNN- control methods, to mitigate the spread of disease and
based approach in accurately classifying leaf images and its ensure the health and productivity of plants.
potential for assisting farmers and agricultural experts in
timely disease diagnosis and management. The proposed
II. EXSISTING SYSTEM
system had a validation accuracy of 98.6% and a training
accuracy of 98.7%. In the realm of agricultural data, automated
Index Terms—Convolutional Neural Network, Smart detection and diagnosis of leaf diseases is greatly sought.
Agriculture, Leaf Disease, Image Processing, Machine Machine learning plays an important role and can handle
Learning difficulties in leaf disease identification rather well. The
existing system presents a novel disease detection approach
I. INTRODUCTION based on machine learning techniques. Various leaf
Agribusiness has developed into a significant source diseases are considered, and different machine learning
of financial improvement, and in India, 80% of the techniques are utilized for their classification. Initially,
population is dependent on agriculture. Depending on the features of leaf disease images are extracted. Then, various
type of soil, local climate, and economic value, the rancher machine learning techniques are applied to classify the
selects the best yield. The majority of the time, farmers images. The existing system achieved an accuracy of
commit suicide as a result of production loss because they 81.8% using the Quadratic SVM classifier. Shape features
are unable to repay the bank debts they have taken out for such as area, roundness, area to lesion ratio, etc., were also
farming. We have seen that the environment is changing utilized to differentiate between different types of leaf
steadily right now, which is bad for the crops and pushes diseases.
farmers into debt and suicide. When different mathematical
or statistical techniques are used to data, these risks can be The downsides of existing framework are measured such
reduced, and by doing so, we can tell the farmer about plant as,
diseases that affect his agricultural area.  While machine learning algorithms can be highly
The agricultural industries started looking for new accurate in identifying leaf diseases, they are not
ways to enhance food production as a result of the infallible.
expanding population, climatic changes, and poor  Factors such as lighting conditions, image quality,
governance. Researchers are working to create new, and variations in leaf appearance due to factors like
capable, and distinctive technologies for producing age or environmental stress can affect the accuracy
outcomes with great efficiency. This study used 54,306 of identification.
photos, including pictures of ill and healthy plant leaves, to  Machine learning models are typically trained on a
train a convolution neural network model to identify 14 specific set of diseases and may struggle to
crop species, 26 diseases, and 38 classes. In a held-out test accurately identify diseases that were not included
set, this trained model has an accuracy of 99.35%. In this in the training dataset.
methodology, plant diseases are classified and detected
using machine learning techniques and image processing III. PROPOSED METHOD
techniques, respectively. Supervised learning and
unsupervised learning are the two main classifications of The main objective of this project is to create a
machine learning. The categorization of agricultural model that can differentiate between healthy and sick
products is done to stop losses in yield and quantity. If the harvest leaves and, in the event that the crop has a disease,
proper evaluation isn't made of this technique or class, it to determine which sickness it is. This study examined
has major effects on plants and affects the quality or plant photos, including those of the tomato, blueberry,
productivity of the corresponding product. Several orange, peach, corn (maize), potato, raspberry, soy bean,
strawberry, and blueberry. The referenced dataset is taken
from the well-known public source kaggle. The classify the disease. Plant diseases pose a significant
Convolutional Neural Network (CNN) is a type of deep challenge as they can cause substantial reductions in both
learning algorithm is used in this project. The CNN is used the quality and quantity of agricultural products. Automatic
here because of its ability to automatically learn detection of plant diseases is crucial for monitoring large
hierarchical representation of features makes CNN crop fields and identifying disease symptoms early. The
particularly effective for image classification tasks. The proposed system provides a software solution for the
initial phase of the system for identifying and classifying automatic detection and classification of plant leaf diseases.
plant diseases is dataset loading. This dataset of plant
photos contains photographs of both healthy and sick The author Omkar Kulkarni[2] (2018) proposes a deep
plants. The second step in the approach for identifying and learning-based model which is trained using public dataset
classifying plant diseases is pre-processing. At this point, containing images of healthy and diseased crop leaves. The
only maintain the relevant data and remove irregular and model serves its objective by classifying images of leaves
noisy data from the dataset. The next step in the process of into diseased category based on the pattern of defect. In
identifying and categorising plant diseases is feature recent times, drastic climate changes and lack of immunity
extraction. Image categorization greatly benefits from in crops has caused substantial increase in growth of crop
feature extraction. There are several uses for feature diseases. Due to rapid growth in variety of diseases and
extraction. This method has discovered that morphological adequate knowledge of farmer, identification and treatment
outcomes provide preferred results over other aspects. The of the disease has become a major challenge. The leaves
dataset can also be used to identify rare plants and rare have texture and visual similarities which attributes for
diseases. The classification technique is the next phase of identification of disease type. Hence, computer vision
the system for recognising and classifying plant infections. employed with deep learning provides the way to solve this
Users can recognise and describe the plant infection at the problem.
final stage. This represents the system's ultimate end stage.
The proposed system had a validation accuracy of 98.6% The author Chittabarni Sarkar[3] (2020) point out the
and a training accuracy of 98.7%. Identification of leaf disorder plays an important role in the
economic prosperity of any country. Many parts of a plant
The advantages of the proposed method are: can be infected by a virus, fungal, bacteria, and other
infectious organisms but here we mainly considered the
detection of leaf disease of a plant as a research topic. We
 Early identification allows for prompt
have performed an in-depth study of this topic from 2010 to
intervention, preventing the spread of diseases
2022 and found that many researchers use multispectral or
and minimizing crop losses.
hyperspectral imaging to study crop diseases. Machine
 This can significantly improve crop yield and
learning (ML) and deep learning (DL) models are used to
quality. The suggested system model is more
classify different types of leaf diseases. We made a workflow
effective.
mechanism to help researchers in this field. Support vector
 The suggested method provides an accurate machine (SVM), Random Forest, and multiple twin SVM
identifying strategy for plant leaf diseases with (MTSVM) are popular ML models for predicting leaf
good accuracy and leverages auxiliary disease, while convolutional neural networks (CNN).
Classifiers as regularizes.
 Reduces labor costs and the need for expensive The author Sunku Rohan, Triveni S Pujar, Suma VR
chemical treatments, enhancing profitability. Amog Shetty, Rishabh F Tated[4] (2019) describes about the
 The suggested method provides an accurate agriculture field has a high impact on our life. Agriculture is
identifying strategy for plant leaf diseases with the most important sector of our Economy. Farmers are
good accuracy and leverages auxiliary difficult to identify the leaf disease so they produce less
Classifiers as Regularizes. production. So that can solve the problem of related to crop
 Accurate disease identification enables farmers disease. Due to the improvement and development in
to select appropriate control measures, such as technology where devices are smart enough to recognize and
specific fungicides or cultural practices, detect plant diseases. In this paper focus on plant disease
tailored to the identified pathogen. detection using image processing techniques. This paper
access open dataset images that consist 5000 images of
 This reduces the risk of overuse of pesticides healthy and diseased plant leaves, and there used semi
and promotes sustainable agricultural supervised techniques for crop types and detect the disease of
practices. four classes.
IV. RELATED WORK The author Omkar Kulkarni (2018) [5] proposes a deep
learning-based model which is trained using public dataset
The author S. Arivazhagan[1] (2020) proposed the color containing images of healthy and diseased crop leaves. The
transformation structure is taken for the input RGB image, model serves its objective by classifying images of leaves
and then by means of a specific threshold value, the green into diseased category based on the pattern of defect. In
pixels are detected and uninvolved which is followed by recent times, drastic climate changes and lack of immunity in
segmentation process, and for obtaining beneficial crops has caused substantial increase in growth of crop
segments the texture statistics are computed. At last, diseases. Due to rapid growth in variety of diseases and
classifier is used for the features that are extracted to adequate knowledge of farmer, identification and treatment
of the disease has become a major challenge. The leaves Processing Module, where a diverse dataset of leaf images is
have texture and visual similarities which attributes for amassed, encompassing both healthy leaves and those
identification of disease type. Hence, computer vision afflicted by various diseases. Each image is meticulously
employed with deep learning provides the way to solve this labelled to denote the corresponding disease or health status.
problem. with database where disease classification images are
already stored and then the result is predicted. And also, the
The author (2019) Karunya Rathan, Abirami Devaraj, system provides Pesticide for the detected leaf disease. The
K Indira and Sarvepalli Jaahnavi [6] talks in this paper, training process entails feeding batches of images into the
agriculture has become far more than simply a method to model, computing the loss (i.e., the disparity between
feed ever growing populations. It’s important wherever in predicted and actual labels), and backpropagating this error
additional than seventieth population of an Asian country is to update the model parameters. To prevent overfitting, the
depends on agriculture. Leaf disease detection may be model's performance is regularly evaluated on a separate
stopping agricultural losses. The aim of this is to develop a validation set, enabling the adjustment of hyperparameters
software system answer that mechanically find and classify to optimize performance.
disease. That consist steps like image acquisition, pre-
processing, Segmentation, extraction and classification are VI. WORKING SYSTEM
involves disease detection. The leaves images are used for
detecting the plant diseases. Therefore, use of image CNN Algorithm
process technique to find and classify diseases in
agricultural. In leaf disease identification using Convolutional
Neural Networks (CNNs), a typical algorithmic approach
The author Balakrishna K Mahesh Rao [7] (2019) talks involves several interconnected layers designed to extract
in this paper, the authors used two methods for and analyse features from input leaf images. The process
identification and classification of healthy and diseased begins with convolutional layers, which serve as feature
tomato leaves. In the first technique, the tomato leaf is extractors by applying filters to detect patterns such as
classified as healthy or unhealthy using the k-nearest edges, textures, and shapes within the images. Stacked
neighbour approach. Later, in the second technique, they convolutional layers progressively capture more complex
classify the unhealthy tomato leaf using probabilistic neural features, enhancing the model's ability to discern subtle
network and the k-nearest neighbour approach. Plant differences between healthy and diseased leaves.
diseases are important in the crops production, which Activation functions like Rectified Linear Unit
affects food security and reduces the profit of farmers. (ReLU) introduce non-linearity, allowing the network to
Identifying the diseases in plants is the key to avoiding learn intricate relationships in the data. Pooling layers
losses by proper feeding measures to cure the diseases early interspersed between convolutional layers down sample the
and avoiding the reduce in production. The features are like feature maps, reducing computational complexity while
GLCM, Gabor, and color are used for classification preserving important information. The data collection and
purposes. pre-processing module in leaf disease identification
involves gathering a diverse dataset of plant leaf images
this undergo pre-processing to enhance quality, feature
V. SYSTEM DESIGN extraction for relevant information.
Gather a dataset of leaf images, including both
The System Architecture briefly defines about the healthy leaves and leaves affected by various diseases.
proposed system work, which helps to make early detection Divide the dataset into training, validation, and test sets.
of leaf disease. Typically, you might use around 70-80% for training, 10-
15% for validation, and 1015% for testing. Convolutional
Neural Networks (CNNs) are commonly used for image
classification tasks and are likely to be effective for this
application.
 Load a dataset from a source.
 Split the dataset into "training" and "test" data.
 Train the CNN in the data.
 Using the classifiers above, predict the labels of the
test data.
 Measure accuracy and see classification.

The Deployment Module for leaf disease identification

employs Flask to create a user-friendly web application. It
includes a visually appealing interface for image upload and
Fig. 4.1. System Architecture integration. The deployment module in leaf disease
The process of leaf disease identification using identification involves making the trained model accessible
Convolutional Neural Networks (CNNs) typically involves for real-world use, typically through a web or windows. If
several distinct modules to ensure a systematic approach. deploying as a windows application, develop the user
The initial phase comprises the Data Collection and Pre- interface to allow users to upload leaf images and receive
predictions for disease identification. Integrate the loading
and inference functionality into the application Load a
dataset from a source. D. Pesticide Recommendation

 Split the dataset into "training" and "test" data.

 Train the classifier in the data.
 Using the classifiers above, predict the labels of the
test data.
 Measure accuracy and see classification.

VII. OUTPUT

A. Home Page

E. Contact

B. Submit Page

VIII. CONCLUSION

In conclusion, the scope of this work says that earlier detection

of leaf disease can be performed and the same will be conveyed
with the respective pesticides to the user through mobile app
with more accuracy. In addition to this, this model can also able
C. Disease Classification to convey about the temperature, soil and humidity level of the
specific area. It is also useful for indoor plantation and terrace
gardening users. This paper has achieved its goal to detect and
recognize various types of plants leaf disease using Digital
Image Processing Techniques and with the classification
algorithm. Creating digital farming systems with new IoT
technology and artificial intelligence should help users or
farmers make decisions on a case-by-case basis.
The future enhancement of this work may be extended to large
number of various crops with the help of Raspberry for
controlling and storing data, which helps to upgrade this
system. With accuracy of 90% of this proposed model the user
can easily detect the disease at early stage and provide pesticide
respectively.
REFERENCES
[1] [CHITTABARNI SARKAR (2020) “Detection
and Classification of Leaf Disease: The topic of plant
leaf disease detection and classification using deep
learning and other advanced techniques” IEEE
Internet Things J 5(6):4409–4417.

[2] S. ARIVAZHAGAN (2020) “Detection of

unhealthy region of plant leaves and classification of
plant leaf disease.” 2nd International Conference on
Data, Engineering and Applications (IDEA).

[3] Sunku Rohan, Triveni S Pujar, Suma VR Amog

Shetty, Rishabh F Tated (2019) “CNN based leaf
disease identification and remedy recommendation
system.” IEEE 25th International Conference on
Parallel and Distributed Systems (ICPADS).

[4] Omkar Kulkarni (2018) “Crop disease detection

using deep learning” In: Lect notes informatics (LNI).
Proc - Ser Gesellschaft Fur Inform 266:79–88.

[5] Karunya Rathan, Abirami Devaraj, K Indira and

Sarvepalli Jaahnavi (2019) “Identification of plant
disease using image processing”.

[6] Balakrishna K Mahesh Rao (2019) “Tomato plant

leaves disease classification”. International
Conference on Data Management, Analytics and
Innovation (ICDMAI). IEEE.
[7] G Kavya, R. M., et al (2016). "Image processing
techniques-based plant disease detection." Int J Adv
Found Res Computer: Pragmatic Rev3.8.
[8] J. Peng, Shaobing, Qiyuan Tang, and Yingbin Zou
(2008) "Current status and challenges of rice
production in China." Plant Production Science 12.1:
3-8.

[9] Jiang, Feng, et al (2020) "Image recognition of

four rice leaf diseases based on deep learning and
support vector machine." Computers and Electronics
in Agriculture 179: 105824.

[10] Izay, A., and L. N. Onyejegbu (2017) "Digital

Image Processing for Detecting and Classifying Plant
Diseases." Circulation in Computer Science 2.11:1-7.

[11] Phadikar, Santanu, and Jaya Sil (2008)"Rice

disease identification using pattern recognition
techniques." 11th International Conference on
Computer and Information Technology. IEEE,2008.

[12] Jian, Zhang, and Zhang Wei (2020) "Support

vector machine for recognition of cucumber leaf
diseases."2nd international conference on advance
computer control. Vol. 5. IEEE.

[13] Barbedo, Jayme Garcia Arnal (2013). "Digital

image processing techniques for detecting, quantifying
and classifying plant diseases." SpringerPlus 2.1: 1-12.