Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Prediction of Rice Diseases Using Convolutional

Neural Network (in Rstudio)
A. Sony
Dept. of CSE
Sreenidhi Institute of Science and Technology

Abstract:- Detection of rice pest and diseases, and

proper management and control of pest infested rice
fields may result to a higher rice crop production. Using
modern technologies, like smart phones, farmers can be
aided in detecting and identifying the type of pests and
diseases found in their rice fields. Convolutional Neural
Network using r language are used to find the diseases
in rice by using images of disease leaves. The disease
images are collected from the UCI Machine Learning
Repository contain the three types of diseases namely
Bacterial Leaf Blight, Brown Spot, Leaf Smut.

Keywords:- Convolutional Neural Network, Rstudio, Paddy

Image Dataset.

I. INTRODUCTION

Rice is the world’s most important food crop and

furnish food for more than half of the population. In the last
20 yrs, 3 lakhs farmers have committed suicides, due to
yield loss. Farmers lose an estimated average of 37% of
their rice crop due to pests and diseases every year. So, Fig 1
identifying, analysing, predicting the disease must be fast to
upgrade production. The diseases are mostly caused by  Blast is caused by “Magnaporthe grisea fungus”.These
Bacteria, Fungus, Virus and others like mosquito, green symptoms can be found on leaves,neck,seeds.Classic
hoppers etc. Mostly occurred bacteria diseases are Bacterial diamond shaped lesions seen on leaves,at the neck
Leaf Blight, foot rot etc. and fungus are blast, brown spot, brown to chocolate colour spots .
narrow brown spot, leaf smut, false smut etc. and Rice
tungro is more occurred virus disease.

Convolutional Neural Network algorithm in r plays a

key role to improve the accuracy for detecting diseases. R
is a programming language which is used for many
statistical modelling and data science tasks, it is
dynamically typed, provides more than 10,000 default
packages.

A. Plant Diseases on Paddy Crop

Paddy is the major crop for many farmers ,there are
mainly two seasons in which paddy is grown, one is rabbi
and other karif.Each season takes six months for
production. Due to the environment changes diseases are
attacking the crop result in yield loss. Below are the major
paddy diseases:

 Bacterial Leaf Blight is mostly occurred diseases caused

by “Xanthomonas Oryzae bacteria”.It occurred mostly
in the area that have weeds, stubbles of infected plants
and high use of nitrogen.More observed symptoms are
Milky dew drops can be observed early in the morning. Fig 2

IJISRT19DEC446 www.ijisrt.com 595

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 False smut is caused by” Ustilaginoidea Virenil fungi”  Rice tungro is an virus disease caused by “RTBV”and
at flowering stage.Individual rice grain transformed into “RTSV”.The major symptom is leaves turn yellow and
a mass of velvety spores. more leafhoppers .

Fig 4

II. ABOUT DATASET COLLECTION

The dataset is collected from the UCI repository and

the information about diseases are gained under the
guidance of DR.Srinivas Prasad who was a scientist in
IIRR (Indian Institute of Rice Research centre in
Rajendranagar).

The link where image dataset collected is:

https://archive.ics.uci.edu/ml/datasets/Rice+Leaf+Diseases

Fig 3 Based on the capacity of my system I have trained 15

images and tested three images for prediction.

Fig 5

IJISRT19DEC446 www.ijisrt.com 596

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
After dividing dataset into training and testing, using III. HOW CNN WORKS
Convolutional Neural Network we are going to predict the
testing sample. In CNN it reduce the parameters this can help to save
lot of time.

Fig 6

ConvNet is a type of feed-forward artificial neural C. Steps involved in Pooling Layer:

network in which the connectivity pattern between its In this layer we shrink the image stack into a smaller
neurons is inspired by the organization of the animal visual size.
cortex.  Pick a window size.
 Pick a stride.
Have following layers:  Walk your window across your filtered images.
 Convolution  From each window take the maximum value.
 ReLU Layer
 Pooling We can add one more layers of Convolution, ReLu
 Fully Connected and Pooling then go to the Fully Connected Layer. Here
we take our filtered images and put them into a single list.
A. Steps Involved in Convolution Layer:
Here we will move the feature /filter to every possible Compare the input image value to the default image
position on the image. values, the value which is near to the default image value
Step1: Line up the feature and the image. then that image belongs to that set.
Step2: Multiple each image pixel by the corresponding
feature pixel. IV. IMPLEMENTATION DIAGRAM
Step3: Add them up and divide by total number of pixels in
the feature.
Step4: Now to keep track of where that feature was, we
create a map and put the value of the filter at that place.
Step5: Similarly we move the feature to every other
positions of the image and will see how the feature matches
that area.

B. Steps Involved in ReLU Layer:

Rectified Linear Unit is an activation function only
activates a node if the input is above a certain quantity,
while the input is below zero, the output is zero. In this
layer we remove every negative values from the filtered
images and replaces it with zeros.

Fig 7

IJISRT19DEC446 www.ijisrt.com 597

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
V. STEPS TO BE PERFORMED

Fig 8

 Load Packages and Read Images:

We require tensor flow as the main package and sub packages are EBImage, keras.
library(tensorflow)
library(keras)
library(EBImage)

Fig 9

 Explore, Resize & Combine:

We resize all the images both train and test into 100 * 100.

IJISRT19DEC446 www.ijisrt.com 598

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig 10

 Reorder Dimension and Response:

By using function “aperm” we change the order of dimensions.

 One Hot Encoding:

Using to_categorical we perform one hot encoding.

Fig 11

 Model:
Here I have used 32 filters, kernel size is 3 * 3 ,activation is ReLu and input shape is 100 * 100 *3.

IJISRT19DEC446 www.ijisrt.com 599

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig 12

Specified learning rate as 0.01, decay is 1e-6 and validation split as 0.2 (it means 20 % data for validation). For better
outcome re-run the model again.

Fig 13

IJISRT19DEC446 www.ijisrt.com 600

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
The graph below shows the loss and accuracy for training and validation data. Accuracy is top one and loss is at bottom.
Accuracy should be larger the better and loss be smaller the better.

Fig 14

 Evaluation & Prediction – Train and Test Data:

The accuracy got is 0.866, if we involve more number of training samples we can improve performance. Here two samples
are correctly predicted.

Fig 15

IJISRT19DEC446 www.ijisrt.com 601

Volume 4, Issue 12, December – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
VI. CONCLUSION

In this paper, particularly focused on identifying the

diseases which occur in paddy using r language .By
improving the training images we achieve better results. In
future, we can also predict disease name and provide
suitable pesticides.

REFERENCES

[1]. Dan, C., Meier, U., Masci, J., Gambardella, L.M.,

Schmidhuber, J. 2011. Flexible, high performance
convolutional neural networks for image
classification. Proceedings of the 22nd International
Joint Conference on Artificial Intelligence, vol. 2, pp.
1237–1242.
[2]. http://refhub.elsevier.com/S0925-2312(17)31138-
4/sbref0002
[3]. jhttp://refhub.elsevier.com/S0925-2312(17)31138-
4/sbref0001
[4]. Hemanta Kalita, Shikhar Kr. Sarma, “Expert System
for Diagnosis of Diseases of Rice Plants: Phototype
Design and Implementation,” 2016
[5]. International Conference on Automatic Control and
Dynamic Optimization Technique (ICACDOT),
International Institute of Information Technology,
Pune, 2016. [2] Chiunnapa Buntaoet, Wongkot
Sriurai. “Diagnosis System for Chili Disease by using
Data Classification Technique,” The Eight National
Conference on Computing and Information
Technology, NCCIT, 2012.
[6]. H. Park, J. S. Eun and S. H. Kim, “Image-based
disease diagnosing and predicting of the crops through
the deep learning mechanism”, In Information and
Communication Technology Convergence (ICTC),
IEEE 2017 International Conference on, pp. 129-131,
2017.
[7]. K. Elangovan and S. Nalini, “Plant disease
classification using image segmentation and SVM
techniques”, International Journal of Computational
Intelligence Research, vol. 13(7), pp. 1821-1828,
2017.
[8]. S. H. Lee, C. S. Chan, S. J. Mayo and P. Remagnino,
“How deep learning extracts and learns leaf features
for plant classification”, Pattern Recognition, vol. 71,
pp. 1-13, 2017.
[9]. K. P. Ferentinos, “Deep learning models for plant
disease detection and diagnosis”, Computers and
Electronics in Agriculture, vol. 145, pp. 311318,
2018.

IJISRT19DEC446 www.ijisrt.com 602