RV College of

Engineering Go, change the world

Go, Change the world

RV College of
Engineering
Department of Master of Computer Applications

20MCA 42 –Technical Seminar– Phase II

Plant Disease Detection Using Pre-Trained

Models

Department of Master of Computer Applications

 RV College of
Engineering Agenda Go, change the world


Introduction


Literature Survey


Summary of Literature Survey


Technical Relevance


Block diagram for the proposed approach/solution


Tools and Technologies


Sustainability and Societal Concern


Conclusion


References Master of Computer Applications
 RV College of
Engineering Introduction Go, change the world

● What is Transfer Learning/Pre-trained models?

● Where can Transfer Learning/Pre-trained models be applied?

● There are various pre-trained models used for disease detection

● Objectives:

1. To understand how pre-trained models and transfer learning is used in machine Learning

2. To have a detailed comparison of various pre-trained models used to detect plant diseases

3. To research and understand which pre-trained model or transfer learning methodology works best

with their corresponding accuracies

Master of Computer Applications
 RV College of
Engineering Introduction Go, change the world

•In India the major percentage around 70% of the population relies on agriculture. In order to produce a
healthy yield, the identification of the plant diseases is important

•The diseases that affect plants cannot be manually observed by the farmers.t's a troublesome and a very
time-consuming process for them to detect the plant diseases manually

•This study mainly concentrates on the transfer learning phenomenon based on different pre-trained models
such as VGG-16, ResNet-34, ResNet-50, and ResNet-50 v2 and then compared these models based on
various standard evaluation metrics such as accuracy, recall and precision.

•The dataset considered for the study includes the various disease and healthy leaves of different plants and
crops such as Apple, Corn, Pepper, Potato, and Tomato, etc
Master of Computer Applications
RV College of
Engineering Literature Review
S No Author and Paper title
1 Bincy Chellapandi; M.
Vijayalakshmi; Shalu Chopra
Title: ”Comparison of Pre-
Trained Models Using Transfer
Learning for Detecting Plant
Disease”
2 Marko Arsenovic,Mirjana
Karanovic,Srdjan
Sladojevic,Andras Anderla
Title: "Solving Current
Limitations of Deep Learning
Based Approaches for Plant
Disease Detection”
Master of Computer Applications
Details of Publication
2021 International Conference
on Computing,
Communication, and Intelligent
Systems (ICCCIS)
Symmetry 2019, 11(7), 939
Go, change the world
Summary of the Paper
In this paper, a deep learning-
based model and transfer
learning-based models to
classifying images of diseased
plant leaves into 38 categories
of plant disease based on its
defect on a Plant Village
dataset. Eight pre-trained. We
found that DenseNet achieves
the best result on the test data
with an accuracy of 99%
In this paper,two approaches
were used to augment the
number of images in the
dataset: traditional
augmentation methods and
state-of-the-art style
generative adversarial
networks.
RV College of
Engineering Literature Review
S No Author and Paper title
3 Krishnaswamy Rangarajan, A., Sci Rep 10, 2322 (2020).
Purushothaman, R.
Title: "Disease Classification in
Eggplant Using Pre-trained
VGG16 and MSVM"
4 Nisar Ahmad; Samayveer
Singh
Title: "Comparative Study of
Disease Detection in Plants
using Machine Learning and
Deep Learning.”
5 Sk Mahmudul hassan and
Arnab Kumar Maji
Title: "Plant Disease
Identification Using a Novel
Convolutional Neural
Network" Master of Computer Applications
Details of Publication
https://doi.org/10.1038/s4159
8-020-59108-x
 2021 2nd International
Conference on Secure Cyber
Computing and
Communications (ICSCCC)
2021 IEEE Fifth Ecuador
Technical Chapters Meeting
(ETCM), 2021, pp. 1-6, doi:
10.1109/ETCM53643.2021.959
0750.
Go, change the world
Summary of the Paper
Pre-trained Visual Geometry
Group 16 (VGG16) architecture
has been used and the images
have been converted to other
color spaces namely Hue
Saturation Value (HSV), YCbCr
and grayscale for evaluation.
In this paper, the machine
learning and deep learning
method like CNN architecture
based on transfer learning by
training them on the available
Plant Village dataset.
In this paper, a novel deep
learning model based on the
inception layer and residual
connection. Depth wise
separable convolution is used
to reduce the number of
parameters
RV College of
Engineering Literature Review
S No Author and Paper title
8 Bipasa Bose; Jyotsna
Priya; Sonam Welekar; Zeyu
Gao
Title: "Hemp Disease Detection
and Classification Using
Machine Learning and Deep
Learning"
9 Sk Mahmudul Hassan , Arnab
Kumar Maji , Michał Jasi,
Zbigniew Leonowicz and
Elzbieta Jasi
Title: "Identification of Plant-
Leaf Diseases Using CNN and
Transfer-Learning Approach"
10 Waleed Albattah, Marriam
Nawaz, Ali Javed, Momina
Masood & Saleh Albahli
Title: "A novel deep learning
method for detection and
classification of plant diseases”
Master of Computer Applications
Details of Publication
2020 IEEE Intl Conf on
Parallel & Distributed
Processing with Applications,
Big Data & Cloud Computing,
Sustainable Computing &
Communications, Social
Computing & Networking
Electronics 2021, 10, 1388.
https://doi.org/10.3390/
electronics10121388
Complex & Intelligent
Systems volume 8, pages507–
524 (2022)
Go, change the world
Summary of the Paper
 The paper uses pre-trained
deep learning ensemble
models with transfer learning.
It reports comparative
evaluation results of the three
deep learning ensemble
models with an SVM-based
model.
The implemented models were
trained with an open dataset
consisting of 14 different plant
species, and 38 different
categorical disease classes and
healthy plant leaves.
The presented method follows
three steps. In the first step,
annotations are developed to
get the region of interest.
Secondly, an improved
CenterNet is introduced in
which DenseNet-77 is
proposed for deep keypoints
extraction.
 RV College of
Engineering Summary of the Literature Survey Go, change the world

•The earliest application of deep learning techniques was in the recognition of plant images using
the vein patterns of leaves
•The benefits of using transfer learning are a decrease in training time, generalization error, and
computational cost of building a DL model
•The machine learning and deep learning methods like CNN architecture based on transfer
learning are been explored by training them on the available Plant Village dataset and these
experiments help us understand the various parameters that contribute to the performance of the
model
•The depth wise separable convolution is used to reduce the number of parameters in deep
learning models to achieve better accuracy in terms of performance of the model
•Multiple feature extraction improves the performance of the model. In a residual network, it has
a shortcut connection that basically feeds the previous layer output to the next layer, which
strengthens features and improves accuracy
Master of Computer Applications
 RV College of
Engineering Technical Relevance Go, change the world

•The dataset considered for the implementation is the “PlantVillage” dataset which includes the
various diseased and healthy leaves of Pepper, Potato, and Tomato, and it is an openly available
dataset through Kaggle

•The proposed framework mainly concentrated on the transfer learning phenomenon based on three
different pre-trained models such as VGG-16, ResNet-50, and ResNet-50 v2, and then compared
the three models based on transfer learning models based on various standard evaluation metrics 

• The images are resized to 256×256 pixels. Data augmentation like shearing, zooming, flipping and
brightness change to increase the dataset size to almost double the original dataset size

•The model is trained for 20 epochs with a batch size by changing the hyper-parameters like
learning rate, batch size, optimizer and pre-trained weights

Master of Computer Applications

 RV College of
Engineering Tools and Technologies Go, change the world

• Machine Learning: Machine Learning is the field of study that gives computers the capability
to learn without being explicitly programmed

• Transfer Learning: Transfer learning (TL) is a research problem in machine learning (ML)
that focuses on storing knowledge gained while solving one problem and applying it to a
different but related problem

• Pre-trained Models: A pre-trained model is a saved network that was previously trained on a
large dataset, typically on a large-scale image-classification task

• Python 3: Python is a general-purpose, high-level, interpreted programming language. With the

usage of extensive indentation, its design philosophy places an emphasis on code readability.
Python uses garbage collection and has dynamic typing.

Master of Computer Applications

 RV College of
Engineering Tools and Technologies Go, change the world

• Tensorflow: TensorFlow is an end-to-end open-source platform that has a comprehensive, flexible

ecosystem of tools, libraries, and community resources that lets researchers push the state-of-the-art
in ML and developers easily build and deploy ML-powered applications

• NumPy: NumPy offers comprehensive mathematical functions, random number generators, linear
algebra routines, Fourier transforms, and more

• Pandas: pandas is a fast, powerful, flexible, and easy-to-use open-source data analysis and
manipulation tool, built on top of the Python programming language

• Scikit-learn: Scikit-learn is a free software machine learning library for the Python programming
language. It features various classification, regression and clustering algorithms including support-
vector machines

• Matplotlib: Matplotlib is a comprehensive library for creating static, animated, and interactive
visualizations in Python
Master of Computer Applications
RV College of
Engineering Block diagram Go, change the world

Master of Computer Applications

 RV College of
Engineering Sustainability and Societal Concern Go, change the world

• Transfer learning helps in saving of resources and improved efficiency when training new
models. It can also help with training models when only unlabelled datasets are available, as the
bulk of the model will be pre-trained

• No need to install any other software or the pre-trained models the user wants to implement.
User also don’t have to set up a model structure

• Google Colab is an open-source tool which is freely available without a commercial license,
which is huge advantage for new developers who are trying to learn this tool and get a hands-on
experience without paying for it

•Python is easy to learn, so users could learn to use it quickly along with better productivity.
Python receives the support of comprehensive and easily interpretable documentation

• Scalability of the Transfer of pre-trained representations can improve sample efficiency and
reduce computational requirements for new tasks
Master of Computer Applications
 RV College of
Engineering Sustainability and Societal Concern Go, change the world

WHAT IS PLANT DISEASE DETECTION USED FOR TODAY?

•Farmers are using image recognition to assess the health or disease of a specific plant or crop
and provide advice on what to do

•Some artificial intelligence (AI) maps the life stages and growth of plants, how these will vary
during their cycle, and what the farmer should do

•Many of the agronomists' usual activities are being replaced by AI, which also enables farmers
to reduce their paperwork and administrative burdens

•The term "precision agriculture" is often used to describe this, which is a management strategy
that "collects, processes, and analyses temporal, spatial, and individual data and combines it
with other information to support management decisions according to estimated variability for
improved resource use efficiency, productivity, quality, profitability, and sustainability of
agricultural production"
Master of Computer Applications
RV College of
Engineering Comparative Study

RESNET 50 vs VGG 16
VGG16 RESNET50

LAYERS 16 LAYERS 50 LAYERS

ETA/EPOCH 26s/25 40s/25

ACCURACY 98.74% 98.84%

LOSS 0.72 0.43

OPTIMIZER RMSPROP SGD(Ir=0.0001,momentum=0.9)

METRICS ACCURACY ACCURACY

Master of Computer Applications

RV College of
Engineering Comparative Study

RESNET 50 vs VGG 16

Parameters considered :

•Pooling layers : In ResNet50 Global average pooling layer is used while in VGG 16 multiple
max pooling layers are used

•Edges and Target Objects : ResNet50 is more sensitive to the edges and target objects

•Speed and Complexity : VGG 16 is slow and complex where as ResNet 50 is faster in training
time

•Vanishing Gradient Problem : As the gradient keeps flowing backward to the initial layers,
this value keeps getting multiplied by each local gradient in VGG 16 resulting in a very small
gradient value where as in ResNet 50 the local gradient is set to 1
Master of Computer Applications
 Comparative Study
RV College of
Engineering

Plant Image
Plant Image
……….

2D Convolution

2D Convolution

2D Convolution

VGG 16 RESNET 50
2D Convolution

2D Convolution

Max Pooling
2D Convolution

………..
……………

Disease Classification
Master of Computer Applications
Disease Classification
 RV College of
Engineering Conclusion Go, change the world

• The proposed framework mainly concentrated on the transfer learning phenomenon based on
three different pre-trained models such as VGG-16, ResNet-50, and ResNet-50 v2, and then
compared the three models based on transfer learning models based on various standard
evaluation metrics. 

• VGG-16-based transfer learning model achieved an accuracy of 98.74%, ResNet-50-based

transfer learning model achieved an accuracy of 98.84%, and ResNet-50 v2-based transfer
learning model achieved an accuracy of 98.21%.

• The dataset considered for the implementation is the “PlantVillage” dataset which includes the
various diseased and healthy leaves of Pepper, Potato, and Tomato, and it is an openly available
dataset through Kaggle.

Master of Computer Applications

 RV College of
Engineering References Go, change the world

[1] Bincy Chellapandi; M. Vijayalakshmi; Shalu Chopra, ”Comparison of Pre-Trained Models Using Transfer Learning for
Detecting Plant Disease”, 2021 International Conference on Computing, Communication, and Intelligent Systems (ICCCIS)

[2] Marko Arsenovic,Mirjana Karanovic,Srdjan Sladojevic,Andras Anderla, "Solving Current Limitations of Deep Learning
Based Approaches for Plant Disease Detection”, Symmetry 2019, 11(7), 939.

[3] Krishnaswamy Rangarajan, A., Purushothaman, R, "Disease Classification in Eggplant Using Pre-trained VGG16 and
MSVM", Sci Rep 10, 2322 (2020).

[4] Nisar Ahmad; Samayveer Singh,"Comparative Study of Disease Detection in Plants using Machine Learning and Deep
Learning.”, 2021 2nd International Conference on Secure Cyber Computing and Communications (ICSCCC).

[5] Sk Mahmudul Hassan and Arnab Kumar Maji,"Plant Disease Identification Using a Novel Convolutional Neural Network" ,
2021 IEEE Fifth Ecuador Technical Chapters Meeting (ETCM), 2021, pp. 1-6.

[6] N Rohan Sai, T Sudarshan Rao, G. L. Aruna Kumari,"Comparative Study on Reliability of Transfer Learning to Classify
Plant-Based Diseases", International Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249-8958 (Online),
Volume-10 Issue-6, August 2021.

[7] Abhinav Sagar,Dheeba Jacob,"On Using Transfer Learning For Plant Disease Detection", 2021 11th International
Conference on Advanced Computer Information Technologies (ACIT), 2021.
Master of Computer Applications
 RV College of
Engineering References Go, change the world

[8] Bipasa Bose; Jyotsna Priya; Sonam Welekar; Zeyu Gao ,"Hemp Disease Detection and Classification Using Machine
Learning and Deep Learning", 2020 IEEE Intl Conf on Parallel & Distributed Processing with Applications, Big Data & Cloud
Computing, Sustainable Computing & Communications, Social Computing & Networking.

[9] Sk Mahmudul Hassan , Arnab Kumar Maji , Michał Jasi, Zbigniew Leonowicz and Elzbieta Jasi "Identification of Plant-Leaf
Diseases Using CNN and Transfer-Learning Approach", Electronics 2021, 10, 1388. https://doi.org/10.3390/
electronics10121388.

[10] Waleed Albattah, Marriam Nawaz, Ali Javed, Momina Masood & Saleh Albahli, "A novel deep learning method for
detection and classification of plant diseases”, Complex & Intelligent Systems volume 8, pages507–524 (2022).

[11] P. Dwivedi, S. Kumar, S. Vijh and Y. Chaturvedi, "Study of Machine Learning Techniques for Plant Disease Recognition
in Agriculture," 2021 11th International Conference on Cloud Computing, Data Science & Engineering (Confluence), Noida,
India, 2021.

[12] L. Li, S. Zhang and B. Wang, "Plant Disease Detection and Classification by Deep Learning—A Review," in IEEE Access,
vol. 9, 2021.

[13] F.-Q. Jiang, C. Li, D.-W. Yu, M. Sun, and E.-B. Zhang, ‘‘Design and experiment of tobacco leaf grade recognition system
based on caffe,’’ J. Chin. Agricult. Mech., vol. 40, no. 1, pp. 126–131, Jan. 2019.
Master of Computer Applications
 RV College of
Engineering References Go, change the world

[14] M.-M. Zhou, ‘‘Apple foliage diseases recognition in Android system with transfer learning-based,’’ M.S. thesis, Dept. Inf.
Eng., Northwest A&F Univ., Yangling, China, 2019.
 
[15] Y. Liu, Q. Feng, and S.-Z. Wang, ‘‘Plant disease identification method based on lightweight CNN and mobile application,’’
Trans. Chin. Soc. Agricult. Eng., vol. 35, no. 17, pp. 194–204, Jun. 2019.

[16] J. G. M. Esgario, R. A. Krohling, and J. A. Ventura, ‘‘Deep learning for classification and severity estimation of coffee leaf
biotic stress,’’ Comput. Electron. Agricult., vol. 169, Feb. 2020, Art. no. 105162.

[17]. Y. Xiong, L. Liang, L. Wang, J. She, and M. Wu, ‘‘Identification of cash crop diseases using automatic image
segmentation algorithm and deep learning with expanded dataset,’’ Comput. Electron. Agricult., vol. 177, Oct. 2020.

[18] M. M. Ozguven and K. Adem, ‘‘Automatic detection and classification of leaf spot disease in sugar beet using deep
learning algorithms,’’ Phys. A, Stat. Mech. Appl., vol. 535, Dec. 2019

[19] B. Wang and D. Wang, ‘‘Plant leaves classification: A few-shot learning method based on Siamese network,’’ IEEE
Access, vol. 7, pp. 151754–151763, 2019

[20] D. Das and C. S. G. Lee, ‘‘A two-stage approach to few-shot learning for image recognition,’’ IEEE Trans. Image Process.,
vol. 29, no. 5,Dec. 2020.
Master of Computer Applications