Leaf Disease Detection Using Deep Learning

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

10 VII July 2022

https://doi.org/10.22214/ijraset.2022.45667
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 10 Issue VII July 2022- Available at www.ijraset.com

Leaf Disease Detection Using Deep Learning


Syed Abdul Rafay1, Srujan Vedula2, Shankar Sunkara3, Sravani Sesetti4,
1, 2, 3, 4
Department of Electronics and Communication Engineering Sagi Rama Krishna Raju Engineering College, Bhimavaram,
AP.

Abstract: Diseases in plants cause major production and economic losses as well as reduction in both quality and quantity of
agricultural products. Now a days, plant diseases detection has received increasing attention in monitoring large field of crops.
Farmers experience great difficulties in switching from one disease control policy to another. The naked eye observation of
experts is the traditional approach adopted in practice for detection and identification of plant diseases. In this project, we study
the need of simple plant leaves disease detection system that would facilitate advancements in agriculture. Early information on
crop health and disease detection can facilitate the control of diseases through proper management strategies. This technique
will improves productivity of crops.
The benefits and limitations of these potential methods are also compared. It includes several steps like image acquisition , image
pre-processing and machine learning algorithms

I. INTRODUCTION
Plant disease, an impairment of the normal state of a plant that interrupts or modifies its vital functions. All species of plants, wild
and cultivated alike, are subject to disease. Although each species is susceptible to characteristic diseases, these are, in each case,
relatively few in number.
The occurrence and prevalence of plant diseases vary from season to season, depending on the presence of the pathogen,
environmental conditions, and the crops and varieties grown. Some plant varieties are particularly subject to outbreaks of diseases
while others are more resistant to them. There are different types of plant diseases. Some of them are very common in many plants.
Some of the most common plant diseases are :
1) Black Rot
2) Powdery Mildew
3) Gray Leaf Spots
4) Common Leaf Rust
5) Northern Leaf Blight
6) Early Blight and Late Blight
7) Bacterial Spots
8) Leaf Scorch
9) Black Measles
10) Mosaic Virus
II. PROBLEM STATEMENT
India is one of the largest producers of agriculture. Diseases in plants results in loss of product and profit. More than eighty percent
of population depends on agriculture. So the main economy of India is based on this profession. In India farmers are very dedicated
to achieve good yield of crops every year. So they shall monitor the well being of crops to ensure they are free from diseases and
parasites. So to diagnose diseases we need efficient technology :CNN.

III. ADVANTAGES
High processing speed and calculation accuracy of detecting various diseases. It reduces human work and can provide better and
efficient solution to prevent crop loss. It reduces cost as well as time because computers are faster than humans. We can use the
same technique to implement for different applications like animal detection, human detection, pest detection in farms etc. by
providing different datasets.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2086
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 10 Issue VII July 2022- Available at www.ijraset.com

IV. BLOCK DIAGRAM

FIG: INCEPTION V3-CONVNET

FIG: VGG19 ARCHITECTURE

V. WORKING PRINCIPLE
Stochastic Gradient Descent with Momentum: It is an iterative algorithm, used for reducing the size and to trace the lowest value
of the function. The algorithm can traversee along the path of steepest descent towards the stable point. Adding a momentum term to
the parameter update is one way to reduce this oscillation. The stochastic gradient descent with momentum update is
θℓ+1=θℓ−α∇E(θℓ)+γ(θℓ−θℓ−1), where γ determines the contribution of the previous gradient step to the current iteration. The
observations cane be allotted by , To train a neural network using the stochastic gradient descent with algorithm, specify solvetask
as 'sgtk'. To specify the initial value of the learning rate α, use the 'Initial' as a naming the point pairs. You can also specify different
learning capabilities for different layers and parameters to get better results we consider Adam algorithm. Adam Optimization :
Adaptive Moment Estimation is an algorithm technique The method is really efficient when working with huge data sets . It
requires less space efficient , it is a combination of the ‘gradient descent with the t’ algorithm and the ‘RMSP’ algorithm. Root
Mean Square Propagation (RMSP): Root mean square propogation or Rprop is an adaptive learning algorithm that tries to improve
AdaGrad Instead of taking the cumulative sum of squared gradients like in AdaGrad, it takes the ‘exponential values of changing
means’. where, Wt = weights at time t Wt+1 = weights at time t+1 αt = learning rate at time t ∂L = derivative of Loss Function ∂Wt
= derivative of weights at time t Vt = sum of square of past gradients. [i.e sum(∂L/∂Wt-1)] (initially, Vt = 0) β = Moving average
parameter (const, 0.9) ϵ = A small positive constant (10-8) Input Arguments for training: • Solver — Solver for training network. •
'Plots' — Plots to display during network training • 'Verbose' — Indicator to display the progress The verbose output displays the
following information: Classification Networks: • Epoch: Epoch number. An epoch corresponds to a full pass of the data. •
Iteration: Iteration frequency A repetition of steps corresponds to a small groups of batches. • Time Elapsed: Time elapsed in hours,
minutes, and seconds. • Mini-batch Accuracy: Classification accuracy on the small groups. • Validation Accuracy: Classification
accuracy on the validation data. If you do not specify validation data, then the function does not display this field. • small group
Loss: Loss on the small group.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2087
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 10 Issue VII July 2022- Available at www.ijraset.com

If the output layer is a Classification Layer object, then the loss is the cross entropy loss for multiclass classification problems with
mutually exclusive classes. • Validation Loss: Loss on the validation data. If the output layer is a Classification Output Layer object,
then the loss is the cross entropy loss for multiclass classification problems with mutually exclusive classes. If you do not specify
validation data, then the function does not display this field. • Base Learning Rate: Base learning rate. The software multiplies the
learn rate factors of the layers by this value. • 'Verbose Frequency' — Frequency of verbose printing. Frequency of verbose printing,
which is the number of iterations between printing to the command window, specified as the comma-separated pair consisting of
'Verbose Frequency' and a positive integer. This option only has an effect when the 'Verbose' value equals true. Mini-Batch Options:
• 'MaxEpochs' — Maximum number of epochs. An iteration is one step taken in the gradient descent algorithm towards minimizing
the loss function using a minibatch. An epoch is the full pass of the training algorithm over the entire training set. • 'Minibatch Size'
— Size of mini-batch. A mini-batch is a subset of the training set that is used to evaluate the gradient of the loss function and update
the weights. • 'Shuffle' — Option for data shuffling. • 'Validpoints' — Data to use for training • 'Validation Frequency' — Frequency
of network validation. • 'Validation Patience' — Patience of validation stopping. • Solver Options: • Initilal — Initial learning rate. •
'Learn' — Option for dropping learning rate during training. • ‘Dropseq' — Number of epochs for dropping the learning rate

VI. RESULTS
A. VGG19 – Result
Input

Output:
Accuracy:91.27%

B. Inception v3-result
Input

Output
Accuracy:92.04%

VII. CONCLUSION
It is crucial to detect the diseases in plants in prior to avoid damage to the crops. So CNN technique is well versed in this task. We
can achieve results with great efficiency with inception V3 architecture.

REFERENCES
[1] Plant Disease Detection Using CNN - Turkish Journal of Computer and Mathematics Education (Vol.12 No.12, 2106-2112)
[2] Plant Disease Identification Using Shallow CNN – Agronomy, MDPI

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2088
International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 10 Issue VII July 2022- Available at www.ijraset.com

[3] Identification of Plant-Leaf Diseases Using CNN and Transfer-Learning Approach – Electronics, MDPI
[4] Recognition of Tomato Late Blight Using DWT and Component Analysis– IJECE, Vol.07 No.01, February 2017
[5] An application of K-means clustering and artificial intelligence in pattern recognition for crop diseases – IPCSIT
[6] Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features – CIGR eJournal 15(1): 211-217
[7] Cotton Leaf Disease Identification using Pattern Recognition Techniques – In Pervasive Computing (ICPC), 2015 International Conference on,pp. 1-6. IEEE,
2015.
[8] Plant Disease Detection Using CNN - ISSN: 2277-3878 (Online), Volume-10 Issue-3, September 2021

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 2089

You might also like