BLDEA’s V P Dr P G Halakatti College of Engineering

and Technology
Department of Information science

Phase 2 Demo On
“ Plant Disease Detection Using Machine Learning”
Presentation by
Akshaykumar Jevaragi (2BL17IS006)
Dayanand Balundagi (2BL17IS015)
Pralhad Khavekar (2BL17IS020)
Vishalkumar Wali (2BL17IS042)

Guide:
Prof. Daneshwari Noola
Professor,
Department of Information science
1
Points to be discussed
 Introduction
 Literature survey
 Existing system
 Problem Definition
 Proposed Work and
methodlogy
 Objectives of work
 Applications
 Pros and Cons
2
 Design
 Data flow diagram
 Modules

3
 Introduction
 Crop diseases are a major threat to food security, but their rapid
identification remains difficult in many parts of the world due to the lack of
the necessary infrastructure.
 The combination of increasing global smartphone penetration and recent
advances in computer vision made possible by deep learning has paved the
way for smart device assisted plant disease diagnosis. 

 Machine learning(ML) is one of the biggest topics in computer science at

the moment, and its many uses make it a topic that will continue to be
researched for a very long time.

4
 Convolutional neural networks (CNN) are a popular realm of machine learning, and
are often used for image classification, as in this project.

 This project is an attempt at using the concepts of neural networks and artificial
intelligence(AI) to create an image classifier which can identify healthy and
diseased plant.

5
Literature survey
Authors & year Goals Future perspective About

[1]Konstantions P. Plant disease detection and The significantly high success 87,848 images, containing 25
Ferentinos diagnosis using simple leaves rate makes the model a very different plants in a set of 58
images of healthy and diseased useful advisory or early warning distinct classes
plants, through deep learning tool, and an approach that could
methodologies be further expanded to support
an integrated plant disease
identification system to operate
in real cultivation conditions.

[2]U.Shruthi; V.Nagaveni; This paper presents the stages of In this survey it observed that Bacteria, Fungus and viral
B.K.Raghavendra general plant diseases detection Convolutional Neural Network diseases of plants.
system and comparative study gives high accuracy and detects
on machine learning more number of diseases of
classification techniques for multiple crops.
plant disease detection

6
[3] Rutu Gandhii; Shubham Image-based classification Almost 40% of the world's Random plant samples
Nimbalkar; Nandita system for identification of crop yield is lost to diseases
Yelamanchili;Surabhi Ponkshe plant diseases and pest infestations.
specifically, there is a need
for establishing a local
dataset to be of use to
Indian farmers.

[4]Shima Ramesh; Use of Random Forest in Using machine learning to train Healthy and diseased samples
Ramachandra Hebbar; identifying between healthy and the large data sets available from random Forest.
Niveditha M; Pooja R.; diseased leaf from the data sets publicly gives us a clear way to
Prasad Bhat created detect the disease present in
plants in a colossal scale.

[5] Sharada P. Mohanty, Training a deep convolutional The approach of training deep 14 crop species and 26 diseases 
David P. neural network to identify crop learning publicly available
species and diseases (or absence image datasets presents a clear
Hughes and Marcel Salathe
thereof).  path toward smartphone-assisted
crop disease diagnosis on a
massive global scale.

7
[6] Ashwin Dhakal , Prof. Dr. The detection of plant Demonstrating the feasibility Bacterial Spot, Yellow Leaf
Subarna Shakya diseases, infected leaves and of this approach in the field of Curl Virus, Late Blight and
crops. Plant Disease helps in the Healthy Leaf
diagnosis and high crop
yielding in future.

[7]Amanda Ramcharan, Kelsee To offer both a cost effective Transfer learning approach for Cassava Disease Detection
Baranowski,PeterMcCloskey, B and scalable technology for image recognition of field
abualiAhmed,James disease detection in Cassava by images offers a fast, affordable,
Legg and David P. Hughes using transfer learning and easily deployable strategy
approach. for digital plant disease
detection in the future.

[8]  Yu Sun, and Jianxin Wang Fine-grained disease severity The proposed deep learning Apple black rot 
classification, as the method model may have great potential
avoids the labor-intensive in disease control for modern
feature engineering and agriculture.
threshold-based segmentation.

8
Existing system
 The existing method for plant disease detection is simply naked eye
observation by experts through which identification and detection of plant
diseases is done. Here the steps followed are,
 Proper plant identification
 Recognizing healthy plant appearance
 Identify characteristic symptoms
 Identify symptom variability
 Laboratory analysis
These above mentioned methods are traditional and so time consuming.
9
Problem Definition
 The symptoms of plant diseases are conspicuous in different parts of a plant
such as leaves.
 In order to detect a plant disease at very initial stage, use of automatic disease
detection technique is advantageous.
 Manual detection of plant disease using leaf images is a tedious job. Hence, it is
required to develop computational methods which will make the process of
disease detection and classification using leaf images automatic.

10
Proposed Work
 In order to develop accurate image classifiers for the purposes of plant disease
diagnosis, we use a large, verified dataset of images of diseased and healthy
plants.
 Then we build CNN(convolutional neural network) model.

 Deep neural networks are trained by tuning the network parameters in such a
way that the mapping improves during the training process.

 Then this model will be measured for its performance and we keep tuning neural
network model to improve its performance by applying machine learning
algorithms. 11
Proposed Methodology

12
 Objectives of work
 Computer vision and object recognition in particular,
has made tremendous advances in the past few years.
 Deep neural networks have recently been successfully applied in
many diverse domains as examples of end to end learning.
 Neural networks provide a mapping between an input—such as
an image of a diseased plant—to an output—such as a
crop~disease pair.

13
Applications
 Farming
 Farmer's economic growth depends on the quality of the products that they
produce, which relies on the plant's growth and the yield they get.
 Therefore, in field of agriculture, detection of disease in plants plays an
instrumental role.

14
References

1.Konstantinos and P.Ferentinos, Deep learning models for plant disease

detection and diagnosis, Computers and Electronics in Agriculture, Volume 145,
Pages 311-318, February 2018.

2. U.Shruthi, V.Nagaveni and B.K.Raghavendra, IEEE, A Review on Machine

Learning Classification Techniques for Plant Disease Detection,
https://doi.org/10.1109/ICACCS.2019.8728415, 06 June 2019.

3. Rutu Gandhii, Shubham Nimbalkar, Nandita Yelamanchili and Surabhi

Ponkshe,IEEE, Plant disease detection using CNNs and GANs as an
augmentative approach, https://doi.org/10.1109/ICIRD.2018.8376321, 2018.

15
4. Shima Ramesh, Ramachandra Hebbar, Niveditha M, Pooja R and Prasad
Bhat, IEEE, Plant Disease Detection Using Machine Learning,
https://doi.org/10.1109/ICDI3C.2018.00017, 2018.

5. Sharada P. Mohanty, David P.Hughes and Marcel Salathe, Using Deep

Learning for Image-Based Plant Disease Detection,
https://doi.org/10.3389/fpls.2016.01419, 22 September 2016.

6. Ashwin Dhakal and Prof. Dr. Subarna Shakya, Image- Based Plant Disease
Detection with Deep Learning, International Journal of Computer Trends and
Technology (IJCTT) – Volume 61 Number 1 – July 2018.

16
7.Amanda Ramsaran, Kelsee Baranowski,PeterMcCloskey,
BabualiAhmed,James Legg and David P. Hughes, Technical advances in Plant
learning, Deep Learning for Image-Based Cassava Disease Detection,
https://doi.org/10.3389/fpls.2017.01852, 27 October 2017

17
DESIGN

18
Periodically images are obtained by remote sensing. RGB values of the
monitored images are extracted and compared with threshold images. If the
threshold is greater or less than given value, histogram analysis and edge
detection techniques are used to identify particular plant diseases.
Different types of crops given as the input to the Training Model. For each
crop, plenty of Healthy and defected crop images are considered. Set the
threshold value for each crop.
Train the model in a such a way that it should take the proper decision for all
types of crops.
Healthy crop images, Infected crop images, different diseases of different
Crops and Pesticides for that diseases are all considered to get trained model.

19
Data warehouse: deals with training datasets. Both healthy and
diseased leaf images are collected. Once the dataset is ready with
healthy and infected image samples, the threshold is extracted for
both aging and for diseases. Periodically images are obtained by
remote sensing.
Data preparation: RGB values of the monitored images are
extracted and compared with threshold images. If the threshold is
greater or less than given value, histogram analysis and edge
detection techniques are used to identify particular plant diseases.

20
• Training Model: Different types of crops given as the input to
the Training Model. For each crop, plenty of Healthy and
defected crop images are considered. Set the threshold value for
each crop.
• Trained Model: Train the model in a such a way that it should
take the proper decision for all types of crops. Fig.1 shows that
Healthy crop images, Infected crop images, different diseases of
different Crops and Pesticides for that diseases are all considered
to get trained model.

21
DATA FLOW DIAGRAM

Disease Feature
Test image Leaf extraction
Detection

Training dataset Leaf diseases

prediction

22
Modules:
Set up TensorFlow Directory and Anaconda Virtual Environment
● Install TensorFlow (skip this step if TensorFlow-GPU 1.5 is already installed) or
TensorFlow-CPU
● The TensorFlow Object Detection API requires using the specific directory
structure provided in its GitHub repository.
Data Collection
● TensorFlow needs hundreds of images of an object to train a good detection
classifier.
● To train a robust classifier, the training images should have random plants in the
image along with the desired plants and should have a variety of backgrounds and
lighting conditions.
● There should be some images where the desired plant is partially obscured,
overlapped with something else, or only halfway in the picture.
23
Data Preprocessing
• With all the pictures gathered, we are going to label the desired
objects in every picture. Using LabelImg (A tool for labeling
images) we label the images in the dataset. 20% of the data
needs to be allocated for testing and 80% data for the training.
• Generate Training Data
We will have to write python scripts to generate training data.
The script helps in generating TFRecord files which will be used
in training the models.

24
Create Label Map and Configure Training
Label map
The label map tells the trainer what each plant is by defining a
mapping of class names to class ID numbers.
Configure training
It defines which model and what parameters will be used for
training.

25
Run the Training

● It will start high and get lower and lower as training

progresses.
● We will perform our training on the Faster-RCNN-Inception-
V2 model. The training keeps reporting the loss at regular
intervals and tries to reduce it step by step. Prediction This is the
final module where the model predicts the input with high
accuracy.

26
 a n k you
Th