agriculture

Review
Review on Convolutional Neural Network (CNN) Applied to
Plant Leaf Disease Classification
Jinzhu Lu 1,2, *, Lijuan Tan 1,2 and Huanyu Jiang 3

1 Modern Agricultural Equipment Research Institute, Xihua University, Chengdu 610039, China;
[email protected]
2 School of Mechanical Engineering, Xihua University, Chengdu 610039, China
3 College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou 310058, China;
[email protected]
* Correspondence: [email protected]

Abstract: Crop production can be greatly reduced due to various diseases, which seriously endan-
gers food security. Thus, detecting plant diseases accurately is necessary and urgent. Traditional
classification methods, such as naked-eye observation and laboratory tests, have many limitations,
such as being time consuming and subjective. Currently, deep learning (DL) methods, especially
those based on convolutional neural network (CNN), have gained widespread application in plant
disease classification. They have solved or partially solved the problems of traditional classification
methods and represent state-of-the-art technology in this field. In this work, we reviewed the latest
CNN networks pertinent to plant leaf disease classification. We summarized DL principles involved
in plant disease classification. Additionally, we summarized the main problems and correspond-
ing solutions of CNN used for plant disease classification. Furthermore, we discussed the future





 development direction in plant disease classification.
Citation: Lu, J.; Tan, L.; Jiang, H.
Keywords: plant disease classification; deep learning; machine learning; convolutional neural network
Review on Convolutional Neural
Network (CNN) Applied to Plant
Leaf Disease Classification.
Agriculture 2021, 11, 707. https://
doi.org/10.3390/agriculture11080707 1. Introduction
The Food and Agriculture Organization of the United Nations (http://www.fao.org/
Academic Editors: Grazia Licciardello publications/sofi/2020/en/, accessed on 5 December 2020) reported that the number of
and Giuliana Loconsole hungry people in the world has been increasing slowly since 2014. Current estimates show
that nearly 690 million people are hungry, and they account for 8.9% of the world’s total
Received: 27 May 2021
population; this figure represents an increase of 10 million in 1 year and nearly 60 million
Accepted: 23 July 2021
in 5 years. Meanwhile, more than 90% of people in the world rely on agriculture. Farmers
Published: 27 July 2021
produce 80% of the world’s food [1]; however, more than 50% of crop production is lost due
to plant diseases and pests [2]. Thus, recognizing and detecting plant disease accurately is
Publisher’s Note: MDPI stays neutral
necessary and urgent.
with regard to jurisdictional claims in
The diverse plant diseases have an enormous effect on growing food crops. An iconic
published maps and institutional affil-
example is the Irish potato famine of 1845–1849, which resulted in 1.2 million deaths [3]. The
iations.
diseases of several common plants are shown in Table 1. Plant diseases can be systematically
divided into fungal, oomycete, hyphomycete, bacterial, and viral types. We have shown
some pictures of plant disease in Figure 1. The pictures in Figure 1 were taken in the
greenhouse of Chengdu Academy of Agriculture and Forestry Sciences. Researchers
Copyright: © 2021 by the authors.
and farmers have never stopped exploring how to develop an intelligent and effective
Licensee MDPI, Basel, Switzerland.
method for plant disease classification. Laboratory test approaches to plant samples, such
This article is an open access article
as polymerase chain reaction, enzyme-linked immunosorbent assay, and loop-mediated
distributed under the terms and
isothermal amplification, are highly specific and sensitive in identifying diseases.
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).

Agriculture 2021, 11, 707. https://doi.org/10.3390/agriculture11080707 https://www.mdpi.com/journal/agriculture

 Agriculture 2021, 11, x FOR PEER REVIEW 2 of 19

Agriculture 2021, 11, 707 2 of 18

Table 1. Common diseases of several common plants.

Major
Table 1. Types
Common of Disease
diseases of several common plants.
Plant Reference
Fungal Bacterial Viral
Major Types of Disease
Plant Downy mildew, powdery KianatReference
et al. (2021) [4],
Fungal Angular
Bacterial spot, brown Mosaic virus, yellow
Viral
Cucumber mildew, gray mold, black Zhang et al. (2019) [5],
Downyspot,
mildew, powdery spot, target spot spot virus Kianat etet al.al.
(2021) [4],[6]
anthracnose Angular spot, brown Mosaic virus, yellow spot Agarwal (2021)
Cucumber mildew, gray mold, black Zhang et al. (2019) [5],
spot, target spot Ricevirus
leaf smut, rice
Rice stripe
spot, blight, false smut, Bacterial leaf blight,
anthracnose Chen etetal.
Agarwal al.(2021)
(2021)[7],[6]
Rice Rice stripe rice
blight, false Bacterial leaf blight, black-streaked
Rice leaf smut, ricedwarf Shrivastava
Chen et al.et(2021) [7], [8]
Rice blast bacterial leaf streak al. (2019)
smut, rice blast bacterial leaf streak black-streakedvirus
dwarf virus Shrivastava et al. (2019) [8]
Leaf spotspot
Leaf disease, rust rust Bacterial
disease, stalk rot,
Bacterial stalk rot, Rough
Rough dwarf disease,
dwarf disease, SunSun et al.
et al. (2021)
(2021) [9],Yu et
[9],
Maize
Maize disease, gray leaf spot
disease, gray leaf spot bacterial leaf streak
bacterial leaf streak crimson leaf disease
crimson leaf disease Yu et al. (2014)
al. (2014) [10] [10]
Early blight, late blight, Bacterial wilt, soft Ferentinos (2018) [11],
Tomato Early leaf
blight, late blight, leaf Bacterial wilt, soft Tomato
rot, yellowyellow
Tomato leaf curl virus
leaf curl
Tomato mold rot, canker Abbas et al. (2018)
Ferentinos (2021) [12][11],
mold canker virus Abbas et al. (2021) [12]

Figure1.1.Leaf
Figure Leafspot
spotin
ineight
eightcommon
commonplants.
plants. We
Wetook
tookthese
thesepictures
picturesin
inthe
thegreenhouse
greenhouseofofChengdu
Chengdu
Academy of Agriculture and Forestry Sciences.
Academy of Agriculture and Forestry Sciences.

However,
However, conventional
conventional field
field scouting
scouting for
for diseases
diseases in
in crops
cropsstill
stillrelies
reliesprimarily
primarilyon on
visual inspection of the leaf color patterns and crown structures. People
visual inspection of the leaf color patterns and crown structures. People observe observe
the symp-the
symptoms of diseases
toms of diseases on leaves
on plant plant leaves
with thewith the eye
naked naked
andeye and diagnose
diagnose plant diseases
plant diseases based on
based on experience, which is time and labor consuming and requires specialized
experience, which is time and labor consuming and requires specialized skills [13]. skillsAt[13].
the
At the same time, the disease characteristics among different crops are also
same time, the disease characteristics among different crops are also different due to the different
due to the
variety varietythis
of plants; of condition
plants; this condition
brings a highbrings
degreeaofhigh degree of
complexity complexity
in the in the
classification of
classification of plant diseases. Meanwhile, many studies have focused on the
plant diseases. Meanwhile, many studies have focused on the classification of plant dis- classification
of plant
eases diseases
based based onlearning.
on machine machineUsing learning. Usinglearning
machine machinemethods
learningtomethods to detect
detect plant dis-
plant diseases is mainly divided into the following three steps: first, using preprocessing
eases is mainly divided into the following three steps: first, using preprocessing tech-
techniques to remove the background or segment the infected part; second, extracting
niques to remove the background or segment the infected part; second, extracting the dis-
the distinguishing features for further analysis; finally, using supervised classification or
tinguishing features for further analysis; finally, using supervised classification or unsu-
unsupervised clustering algorithms to classify the features [14–17]. Most machine learning
pervised clustering algorithms to classify the features [14–17]. Most machine learning
studies have focused on the classification of plant diseases by using features, such as the
studies have focused on the classification of plant diseases by using features, such as the
texture [18], type [19], and color [20] of plant leaf images. The main classification methods
texture [18], type [19], and color [20] of plant leaf images. The main classification methods
include support vector machines [19], K-nearest neighbor [20], and random forest [21]. The
include support vector machines [19], K-nearest neighbor [20], and random forest [21].
major disadvantages of these methods are summarized as follows:
The major disadvantages of these methods are summarized as follows:
Low performance [22]: The performance they obtained was not ideal and could not
be used for real-time classification.
Agriculture 2021, 11, 707 3 of 18

Professional database [23]: The datasets they applied contained plant images that
were difficult to obtain in actual life. In the case of PlantVillage, the dataset was taken in an
ideal laboratory environment, such that a single image contains only one plant leaf and the
shot is not influenced by the external environment (e.g., light, rain).
Rarely used [24,25]: They often need to manually design and extract features, which
require research staff to possess professional capabilities.
Requiring the use of segmented operation [26]: The plants must be separated from
their roots to gain research datasets. Obviously, this operation is not good for real-
time applications.
Most of the traditional machine learning algorithms were based on laboratory con-
ditions, and the robustness of the algorithms is insufficient to meet the needs of practical
agricultural applications. Nowadays, deep learning (DL) methods, especially those based
on convolutional neural networks (CNNs), are gaining widespread application in the
agricultural field for detection and classification tasks, such as weed detection [27], crop
pest classification, and plant disease identification [28]. DL is a research direction of ma-
chine learning. It has solved or partially solved the problems of low performance [22],
lack of actual images [23], and segmented operation [26] of traditional machine learning
methods. The important advantage of DL models are that they can extract features without
applying segmented operation while obtaining satisfactory performance. Features of an
object are automatically extracted from the original data. Kunihiko Fukushima introduced
the Neocognitron in 1980, which inspired CNNs [29]. The emergence of CNNs has made
the technology of plant disease classification increasingly efficient and automatic.
The main works of this study are given as follows: (1) we reviewed the latest CNN
networks pertinent to plant leaf disease classification; (2) we summarized DL principles
involved in plant disease classification; (3) we summarized the main problems and corre-
sponding solutions of CNN used for plant disease classification, and (4) we discussed the
direction of future developments in plant disease classification.

2. Deep Learning
DL is a branch of machine learning [30] and is mainly used for image classification,
object detection [31–34], and natural language processing [35–37].
DL is an algorithm based on a neural network for automatic feature selection of data. It
does not need a lot of artificial feature engineering. It combines low-level features to form
abstract high-level features for discovering distributed features and attributes of sample
data. Its accuracy and generalization ability are improved compared to those of traditional
methods in image recognition and target detection. Currently, the main types of networks
are multilayer perceptron, CNN, and recurrent neural network (RNN). CNN is the most
widely used for plant leaf disease classification. As for other DL networks, such as fully
convolutional networks (FCNs) and deconvolutional networks, they are usually used for
image segmentation [38–41] or medical diagnosis [42,43] but are not used for plant leaf
disease classification. CNN usually consists of convolutional, pooling, and fully connected
layers. The convolutional layer uses the local correlation of the information in the image to
extract features. The process of convolution operation is shown in Figure 2. A kernel is placed
in the top-left corner of the image. The pixel values covered by the kernel are multiplied
with the corresponding kernel values, and then the products are summated, and the bias is
added at the end. The kernel is moved over by one pixel, and the process is repeated until all
possible locations in the image are filtered, which is shown in Figure 2. The pooling layer
selects features from the upper layer feature map by sampling and simultaneously makes the
model invariant to translation, rotation, and scaling. The commonly used one is maximum
or average pooling. The process of the pooling operation is shown in Figure 3. Maximum
pooling is to divide the input image into several rectangular regions based on the size of the
filter and output the maximum value for each region. As for average pooling, the output
is the average of each region. Convolutional and pooling layers often appear alternately in
applications. Each neuron in the fully connected layer is connected to the upper neuron, and
Agriculture 2021, 11, x FOR PEER REVIEW 4 of 19

region. As for average pooling, the output is the average of each region. Convolutional
Agriculture and pooling
2021, 11, 707 layers often appear alternately in applications. Each neuron in the fully con- 4 of 18
region. As
nected layer is connected for upper
to the averageneuron,
pooling,and
the the
output is the average offeatures
multidimensional each region. Convolutional
are in-
and pooling layers often appear alternately in applications. Each neuron in the fully con-
tegrated and converted into one-dimensional features in the classifier for classification or
nected layer is connected to the upper neuron, and the multidimensional features are in-
detection tasks [44].
tegrated and converted
the multidimensional into one-dimensional
features features
are integrated and in the
converted classifier
into for classification
one-dimensional features or
in
detection tasks
the classifier for[44].
classification or detection tasks [44].
5×5×1
5×5×1 3×3×1
0 1 2 3 4 3×3×1
0 1 2 3 4 3×3×1 3×3×1
5 6 7 8 9
5 6 7 -18 09 1 7 7 7
-1 0 1 7 7 7
10 11 12 13 14
10 11 12 -1
13 014 1 7 7 7
-1 0 1 7 7 7
15 16 17 18 19
15 16 17 -1
18 019 1 7 7 7
-1 0 1 7 7 7
20 21 22 23 2024 21 22 bias=1,stride=1
23 24
bias=1,stride=1
Input Input Filter Filter Output Output
Figure 2. The processFigure
of convolution
Figure 2.
2. The operation.
The process
process of
of convolution
convolution operation.
operation.

4×4×1 4×4×1 4×4×1 4×4×1

0 1 2 3 2×2×1 0 1 2 3 2×2×1
0 1 2 3 2×2×1 0 1 2 3 2×2×1
4 5 6 7 2×2 filter 5 7 4 5 6 7 2×2 filter 2.5 4.5
4 5 6 7 2×2 filter 5 7 4 5 6 7 2×2 filter 2.5 4.5
8 9 10 11 stride=2 13 15 8 9 10 11 stride=2 10.5 12.5
8 9 10 11 stride=2 13 15 8 9 10 11 stride=2 10.5 12.5
12 13 14 15 12 13 14 15
12 13 14 15 Output
12 13 14 15
Input Output Input Output
Input Input Output
(a) Maximum pooling (b) Average pooling

(a) Maximum pooling (b) Average pooling

Figure 3. The process of pooling operation.
Figure 3. The process of pooling operation.
Figure 3. The process of pooling operation.
For classification tasks, various CNN-based classification models have been devel-
For classification tasks, various CNN-based classification models have been developed
oped in DL-related research, including AlexNet, VGGNet, GoogLeNet, ResNet, Mo-
For classification tasks, various
in DL-related CNN-based
research, classification
including AlexNet, VGGNet, models have been
GoogLeNet, ResNet, devel-
MobileNet, and
bileNet, and EfficientNet. AlexNet [45] was proposed in 2012 and was the champion net-
oped in DL-relatedEfficientNet.
research, including
AlexNet [45] AlexNet,
was proposed VGGNet, in 2012 GoogLeNet,
and was the ResNet,
championMo- network in the
work in the ILSVRC-2012 competition. This network contains five convolutional layers
ILSVRC-2012
bileNet, and EfficientNet. AlexNet competition. This
[45] waslayers. network
proposed contains
in 2012 five convolutional layers and three fully
and three fully connected AlexNet has and was the champion
the following four highlights: net- (a) it is the
connected
work in the ILSVRC-2012 layers.
competition. AlexNet has the following four highlights: (a) it is the first model to use a
first model to use a GPUThis network
device for networkcontains five convolutional
acceleration layers
training; (b) rectified linear units
GPU device for network
and three fully connected acceleration training; (b)four rectified linear units (a) (ReLUs)
it is thewere used as
(ReLUs) layers.
were used AlexNet has the following
as the activation function; (c) local highlights:
response normalization was used;
the activation function; (c) local response normalization was used; (d) in the first two layers
first model to use a(d)GPU device
in the first for
twonetwork
layers of acceleration
the fully connected training; (b)the
layer, rectified
dropout linear unitswas used to
operation
of the fully connected layer, the dropout operation was used to reduce overfitting. Then,
(ReLUs) were usedreduce
the theoverfitting.
as deeperactivation Then, the deeper
networksfunction;
appeared,(c) local
such
networks
as response
appeared,
VGG16, VGG19,
such as VGG16,
normalization
GoogLeNet. wasThese VGG19, Goog-
used; networks use
(d) in the first two LeNet.
layers These
of the networks
fully use smaller
connected stacked
layer, the kernels but
dropout have lower
operation
smaller stacked kernels but have lower memory during inference [46]. Later, was memory
used toduring infer-
researchers
ence
reduce overfitting. found [46].
Then, that Later,
the deeper researchers
networks found that
appeared, when the number of layers of a deep CNN reached
when the number of layers of sucha deep asCNN
VGG16, VGG19,
reached Goog-
a certain depth, blindly
LeNet. These networksaincreasing
certain
usedepth,
smaller blindly
stacked
the number increasing
kernels
of layers the
but
would number
have of layers
lower
not improve would
memory
the not
duringimprove
classification infer-the classifica-
performance but
tion
would performance
cause the but would
network to cause
convergethe network
more to
slowly converge
[47,48]. more
Until slowly
2015, [47,48]. Until
Microsoft lab
ence [46]. Later, researchers found that when the number of layers of a deep CNN reached
2015, Microsoft
proposed lab
the ResNet proposed the
networkofand ResNet network
wonwouldthe first and won the first place in the classifica-
a certain depth, blindly increasing the number layers notplace in thethe
improve classification
classifica- task of the
tion task ofcompetition.
ImageNet the ImageNetThe competition.
network The network
creatively creatively
proposed proposed
residual blocksresidual blocks
tion performance but and would
shortcut cause the network
connections [49], whichto converge
solves the more slowly
problem of [47,48].
gradient Untiland or
elimination
shortcut
gradi-
connections [49], which solves the problem of gradient elimination or gradient explosion,
2015, Microsoft labentproposed
explosion, themaking
ResNetit network
possible toand
buildwon a the first
deeper place model.
network in the classifica-
ResNet influenced the
making it possible to build a deeper network model. ResNet influenced the development
tion task of the ImageNet
development competition.
direction The
of DLnetwork
in academia creatively
and proposed
industry in residual
2016.
direction of DL in academia and industry in 2016. MobileNet was proposed by the Google MobileNet blockswas proposed
and shortcut connections
by the in
teams [49],
2017which
Google teams
and solves
was 2017the
indesigned andproblem
was
for ofand
designed
mobile gradient elimination
for mobile
embedded and embedded
vision or gradi-
applications vision
[50].applica-
In 2019,
ent explosion, makingtions it[50].
the Google Inteams
possible 2019, the Google
to proposed
build teamsnetwork
a deeper
another proposed
outstanding another
model. outstanding
ResNet
network: influenced
EfficientNet network: EfficientNet
theThis
[51]. network
development direction uses of DL in yet
a simple academia and industry
highly efficient compoundin 2016. MobileNet
coefficient was proposed
to uniformly scale all dimensions
by the Google teams of in
depth/width/resolution,
2017 and was designed which
for will not arbitrarily
mobile and embeddedscale the dimensions
vision applica-of the network
as in traditional methods. As for plant disease classification
tions [50]. In 2019, the Google teams proposed another outstanding network: EfficientNettasks, it is not necessary to use
Agriculture 2021, 11, x FOR PEER REVIEW 5 of 19

[51]. This network uses a simple yet highly efficient compound coefficient to uniformly
Agriculture 2021, 11, 707 5 of 18
scale all dimensions of depth/width/resolution, which will not arbitrarily scale the dimen-
sions of the network as in traditional methods. As for plant disease classification tasks, it
is not necessary to use deep networks, because simple models, such as AlexNet and
VGG16, can meetbecause
deep networks, the actual accuracy
simple models,requirements.
such as AlexNet and VGG16, can meet the actual
The DL
accuracy model can be realized using programming languages, such as Python, C/C++.
requirements.
The open-source
The DL model DLcan framework
be realized provides a series of application
using programming languages, programming
such as Python, interfaces,
C/C++.
supports model design,
The open-source assists inprovides
DL framework network adeployment, and avoids
series of application code duplication
programming [52].
interfaces,
At present,
supports DL frameworks,
model design, assists such as PyTorch
in network (https://pytorch.org/,
deployment, and avoids codeaccessed on 5 March
duplication [52].
2021), Tensorflow
At present, (https://www.tensorflow.org/,
DL frameworks, accessed on 7 accessed
such as PyTorch (https://pytorch.org/, March 2021), Cafe
on 5 March
(https://caffe.berkeleyvision.org/,
2021), Tensorflow (https://www.tensorflow.org/, accessed on 8 March 2021),onand
accessed Keras 2021),
7 March (https://keras.io/,
Cafe (https:
//caffe.berkeleyvision.org/,
accessed on 10 March 2021) areaccessed on 8 March 2021), and Keras (https://keras.io/,
widely used.
accessed on 10 increase
The rapid March 2021) of DL areiswidely used. from the widespread development of GPU.
inseparable
The rapid increase
The implementation of DL
of deep CNNis inseparable
requires GPUs from totheprovide
widespread development
computing of GPU.
power support,
The implementation of deep CNN requires GPUs to provide computing
otherwise it will cause the training process to be quite slow or make it impossible to train power support,
otherwise
CNN models it will cause
at all. the training
At present, process
the most usedto be quite slow
is CUDA. When or make
NVIDIA it impossible
launched to CUDAtrain
CNN modelsUnified
(Computing at all. At present,
Device the most used
Architecture) andisAMDCUDA. When NVIDIA
launched Stream, GPU launched CUDA
computing
(Computing
started [46], and Unified
now,Device
CUDA Architecture)
is widely usedand AMD launched Stream, GPU computing
in DL.
started
Image[46],classification
and now, CUDA is widely
is a basic taskused in DL. vision. It is also the basis of object
in computer
Image classification is a basic task
detection, image segmentation, image retrieval, and other in computer vision. It is also the
technologies. Thebasis
basicof object
process
detection, image segmentation, image retrieval, and other technologies.
of DL is shown in Figure 4, taking the task of classification of diseases on the surface of The basic process
of DLgourd
snake is shown in Figure
leaves 4, takingInthe
as an example. task 4,
Figure ofweclassification
use a CNN-basedof diseases on the surface
architecture to extract of
snake gourd leaves as an example. In Figure 4, we use a CNN-based
features, which mainly include convolutional, max-pooling, and full connection layers. architecture to extract
features,
The which mainly
convolutional layer include
is mainly convolutional,
used to extract max-pooling, and full
features of snake gourdconnection
plant leaflayers.
im-
The convolutional layer is mainly used to extract features of
ages. The shallow convolutional layer is used to extract some edge and texture infor- snake gourd plant leaf images.
The shallow
mation, convolutional
the middle layer islayer
usedistoused to extract
extract complex some edge and
texture and texture
part of information,
semantic infor- the
middle layer is used to extract complex texture and part of
mation, and the deep layer is used to extract high-level semantic features. The convolu- semantic information, and
the deep
tional layer layer is used to
is followed byextract high-level
a max-pooling semantic
layer, whichfeatures.
is used to The convolutional
retain the importantlayerin- is
followed by a max-pooling layer, which is used to retain the important
formation in the image. At the end of the architecture is a classifier, which consists of full information in
the image. At the end of the architecture is a classifier, which consists of full connection
connection layers. This classifier is used to classify the high-level semantic features ex-
layers. This classifier is used to classify the high-level semantic features extracted by the
tracted by the feature extractor.
feature extractor.

Figure 4.
Figure Convolutional neural
4. Convolutional neural networks
networks for
for snake
snake gourd
gourd leaf
leaf disease
disease classification.
classification.

In Figure 4, we input a batch of images into the feature extraction network to extract
In Figure 4, we input a batch of images into the feature extraction network to extract
the features and then flatten the feature map into the classifier for disease classification.
the features and then flatten the feature map into the classifier for disease classification.
This process can be roughly divided into the following three steps.
This process can be roughly divided into the following three steps.
1. Step
1. Step1.1.Preparing
Preparingthe
theData
Dataand
andPreprocessing
Preprocessing
2. Step 2. Building, Training, and Evaluating
2. Step 2. Building, Training, and Evaluating the theModel
Model
3. Step 3. Inference and Deployment
 Agriculture 2021, 11, 707 6 of 18

2.1. Data Preparation and Preprocessing

Data are important for DL models. The results are bound to be inaccurate no matter
how complex and perfect our model is as long as the quality of the input data is poor. The
typical percentages of the original dataset intended for training, validation, and test are
70:20:10, 80:10:10, and 60:20:20.
A DL dataset is usually composed of a training set, a validation set, and a test set. The
training set is used to make the model learn, and the validation set is usually used to adjust
hyperparameters during training. The test set is the sample of data that the model has not
seen before, and it is used to evaluate the performance of the DL model. We collected some
public plant datasets from the two websites Kaggle (https://www.kaggle.com/datasets,
accessed on 12 February 2021) and BIFROST (https://datasets.bifrost.ai/, accessed on
15 February 2021), which can be used for detection or classification tasks, as shown in
Table 2. In the literature of DL techniques applied to plant disease classification, the most
used public datasets are PlantVillage [53–55] and Kaggle [56]; notably, many authors also
collect their own datasets [57–60].

Table 2. Some public plant datasets from Kaggle and BIFROST.

Name Number of Images Classes Task Type of View Source

New Plant Diseases Dataset 87,000 38 Image classification Field data Kaggle
PlantVillage Dataset 162,916 38 Image classification Uniform background Kaggle
Flowers Recognition 4242 4 Image classification Field data Kaggle
Plant Seedings Dataset 5539 12 Target detection Field data BIFROST
Weed Detection in Soybean Crops 15,336 4 Target detection Uniform background Kaggle

For snake gourd leaf disease classification, we need a large number of leaf images
of different disease categories. Meanwhile, the disease image data of each category were
roughly balanced. If one disease with a particularly large number of image data is consid-
ered, then the neural network will be biased toward this disease. Apart from sufficient data
on category balance, it also needs data to preprocess including image resize, random crop,
and normalization. The shape of the data varies according to the framework used. Figure 5
shows the tensor shape of the input for the neural network, where H and W represent the
height and width of the preprocessed image, C represents the number of image channels
(gray or RGB), and N represents the number of images input to the neural network
Agriculture 2021, 11, x FOR PEER REVIEW 7 of in
19 a
training session.

Figure
Figure5.5.The
Thetensor shape
tensor ofof
shape the input
the inputneural
neuralnetwork
networkininPyTorch.
PyTorch.

2.2. Building Model Architecture, Training, and Evaluating the Model

Before training, a suitable DL model architecture is needed. A good model architec-
ture can result in more accurate classification results and more rapid classification speed.
Currently, the main network types of DL are CNN, RNN, and generative adversarial net-
Agriculture 2021, 11, 707 7 of 18

2.2. Building Model Architecture, Training, and Evaluating the Model

Before training, a suitable DL model architecture is needed. A good model architecture
can result in more accurate classification results and more rapid classification speed. Cur-
rently, the main network types of DL are CNN, RNN, and generative adversarial networks
(GAN). Among various works, CNN is the most widely used feature extraction network
for the task of plant disease detection and classification [55,61–65].
After the model architecture is established, different hyperparameters are set for
training and evaluation. We can set some parameter combinations and use the grid search
method to iterate through them to find the best one. When training the neural network,
training data are placed into the first layer of the network, and each neuron updates
the weight of the neuron through back-propagation according to whether the output is
equal to the label. This process is repeated until new capability is learned from existing
data. However, whether the trained model has learned new capabilities is unknown. The
performance of the model was evaluated by criteria, such as accuracy, precision, recall, and
F1 score. The concept of a confusion matrix must be introduced first prior to introducing
these indexes specifically. The confusion matrix shows the predicted correct or incorrect
results in binary classification. It consists of four elements: true positive (TP, correctly
predicted positive values), false positive (FP, incorrectly predicted positive values), true
negative (TN, correctly predicted negative values), and false negative (FN, incorrectly
predicted negative values). Then, the accuracy can be calculated as follows:

TP + TN
Accuracy = (1)
TP + FP + TN + FN
Among all the positives predicted by the model, precision predicts the proportion of
correct predictions.
TP
Precision = (2)
TP + FP
Among all real positives, recall predicts the correct proportion of positives [66].

TP
Recall = (3)
TP + FN
The F1 value considers precision (P) and recall (R) rates.

2 2×P×R
F1 = 1 1
= (4)
P + R
P+R

In the studies on plant disease classification, accuracy is the most common evaluation
index [53,60,64,67,68]. Larger values of accuracy, precision, and recall are better. Within
a certain range, when the value of the F1 score is smaller, the better the generalization
performance of the trained model is. When the training and evaluation are complete, the
trained model has a new capability; then, this capability is applied to new data.

2.3. Inference and Deployment

The inference is the capability of the DL model to quickly apply the learning capability
by the trained model to new data and quickly provide the correct answer based on data that
it has never seen [69]. After the training process is completed, the networks are deployed
into the field for inferring a result for the provided data, which they have never seen
before. Only then can the trained deep learning models be applied in real agricultural
environments. We can deploy the trained model to the mobile terminal, cloud, or edge
devices, such as by using an application on the mobile phone to take photos of plant leaves
and judge diseases [70]. In addition, in order to use the trained model better in the field,
the generalization ability of the model needs to be improved, and we can continuously
update the models with the new labeled datasets to improve the generalization ability [71].
Agriculture 2021, 11, 707 8 of 18

3. Problems and Solutions

Before 2015, no notable breakthrough was obtained in plant disease classification.
With the fast development of DL since 2015, DL has been widely used in plant disease
detection and classification and represents state-of-the-art technology in this field. For
plant leaf disease classification, CNN-based models are the most used. In this section, we
introduce and summarize the problems and solutions existing in the development of CNN-
based DL methods applied to plant disease detection and classification. The problems are
caused by extrinsic and intrinsic factors. Sections 3.1 and 3.2 discuss extrinsic factors, and
Sections 3.3 and 3.4 describe intrinsic factors.

3.1. Insufficient Datasets

The most important problem of CNN-based DL’s application of plant disease classifi-
cation is insufficient datasets in size and diversity. All the other introduced problems are
also partially due to this condition.
Mohanty et al. tested the classic network models AlexNet and GoogLeNet with a
public database of 54,306 images collected under controlled conditions to identify 14 crop
species and 26 diseases. They obtained a top accuracy of 99.35%, which demonstrates the
feasibility of this method. However, the accuracy of the model was greatly reduced when
it was tested on a set of images taken under conditions different from the images used for
training because of the insufficient diversity of the training set. In addition, plant disease
identification in this experiment was realized under ideal conditions, such as single leaves,
facing up, in a homogeneous background; thus, the accuracy rate would be much lower in
practical applications [53]. Fuentes et al. aimed to introduce a robust DL-based detector for
real-time tomato disease and pest recognition. All images of plant diseases and pests were
taken in-place, including background variations, different illumination conditions, and
multiple sizes of objects. The precision would be lower in practical application due to the
insufficient number of samples [72]. Sufficient datasets have an important influence on the
practical application. However, collecting data is easily affected by environmental factors,
such as season and climate, and image labeling is also a time-consuming and laborious
task. These factors make producing an effective dataset extremely difficult. Currently, five
ways, namely, transfer learning, data augmentation techniques, few-shot learning, citizen
science, and data sharing, can be used to resolve dataset problems.
Transfer learning is a machine learning technique, where the attained capability from
the previous task is transferred to later tasks [36]. Only a few layers of pretrained networks
are retrained with the new databases, which is good for reducing the need for masses of
datasets [73]. Mukti et al. utilized a transfer learning model based on ResNet50 to recognize
plant diseases. Their dataset contains 87,867 images. A total of 80% of the dataset was used
for training and 20% for validating. The highest accuracy they attained was 99.80% [1].
Coulibaly et al. proposed an approach using transfer learning to recognize mildew diseases
in pearl millet. This approach was based on a classical CNN model VGG16 and pretrained
on public dataset ImageNet. The experiment resulted in a satisfactory performance with
an accuracy of 95% and a recall of 94.5% [74]. Abdalla et al. used three transfer learning
methods for semantic segmentation of oilseed rape images; the experiment resulted in an
accuracy of 96% and demonstrated that transfer learning gained high performance in this
segmentation task [75]. Chen et al. proposed a DL architecture named INC-VGGN, which
utilized the transfer learning by modifying the pretrained VGGNet for the identification
task of plant leaf diseases. The proposed model achieved an accuracy of 91.83% on the
public dataset PlantVillage and 92.00% on their own dataset [60]. Table 3 summarizes some
studies that used transfer learning technology for classification or detection tasks.
Agriculture 2021, 11, 707 9 of 18

Table 3. Studies on transfer learning technology applied to the identification task.

Pretrained Model Dataset Number of Class Best Accuracy Reference

ResNet50 PlantVillage (extended) 38 99.80% Mukti and Biswas (2019) [1]
VGG16 Millet crop images (own) 7 95.00% Coulibaly et al. (2019) [74]
VGG16 Plant images (own) 93.00% Abdalla et al. (2019) [75]
VGGNet ImageNet 9 91.83% Chen et al. (2020) [60]
ResNet-101 NBAIR (extended) 40 95.02% Thenmozhi et al. (2019) [76]
AlexNet ImageNet (partial) 2 98.00% Suh et al. (2018) [77]

Agriculture 2021, 11, x FOR PEER REVIEWThe data augmentation technologies can efficiently increase the number of datasets.
10 of 20
We show some traditional image data augmentation methods, such as rotation, mirror sym-
metry, and adjusting saturation in Figure 6. We have learned some newest augmentation
technologies: AugMix [78], population-based augmentation [79], Fast AutoAugment [80],
RandAugment [81], and CutMix [82].

(a) Original (b) Rotate 45° (c) Rotate 135° (d) Adjust brightness (e) Add gaussian noise

(f) Mirror (g) Texture enhancement (h) Adjust saturation (i) Adjust sharpness (j) Histogram linearization

Figure 6. Traditional
Figure 6. Traditionalimage
imagedata
dataaugmentation
augmentationmethods.
methods.

Liu
Liu et
et al.
al. used
used data
data augmentation
augmentation technologies
technologies to to solve
solve the
the problem
problem ofof insufficient
insufficient
apple
apple pathological images for the identification of four apple leaf diseases. The researchers
pathological images for the identification of four apple leaf diseases. The researchers
used direction disturbance (rotation transformation and mirror symmetry), light distur-
used direction disturbance (rotation transformation and mirror symmetry), light disturb-
bance, and principal component analysis jittering to disturb natural images. With the
ance, and principal component analysis jittering to disturb natural images. With the ap-
application of these image processing technologies, the dataset expanded from 1053 images
plication of these image processing technologies, the dataset expanded from 1053 images
to 13,689 images, and the accuracy with the expanded database improved 10.83% over that
to 13,689 images, and the accuracy with the expanded database improved 10.83% over
in the nonexpanded database [83]. The researchers in [58] used three augmentation meth-
that in the nonexpanded database [83]. The researchers in [58] used three augmentation
ods (noise addition, color jittering, and radial blur) to increase the number of databases.
methods (noise addition, color jittering, and radial blur) to increase the number of data-
Douarre et al. used a novel data augmentation strategy, namely, plant canopy simulation, to
bases. Douarre et al. used a novel data augmentation strategy, namely, plant canopy sim-
generate new annotated data for the segmentation task of plant disease. The results showed
ulation, to generate new annotated data for the segmentation task of plant disease. The
that simulated data had increasing segmentation performance [84]. Table 4 summarizes
results showed that simulated data had increasing segmentation performance [84]. Table
some studies on using data augmentation technologies to expand the dataset.
4 summarizes some studies on using data augmentation technologies to expand the da-
Another method is few-shot learning (FSL), which needs small training sets but with
ataset.
small drop in accuracy. Argüeso et al. [85] introduced FSL algorithms for plant disease
Anotherto
classification method
addressis the
few-shot
problemlearning (FSL), which
of requiring needs smallimage
large annotative training sets but
datasets forwith
DL
a small drop in accuracy. Argüeso et al. [85] introduced FSL algorithms
methods. They split the 54,303 images of the PlantVillage dataset into a source and a for plant disease
classification
target domain.toFirst,
address
theythe problem
used of requiring
the fine-tuning large annotative
Inception V3 networkimage
in the datasets for DL
source domain
methods. They split the 54,303 images of the PlantVillage dataset into a source
to learn general plant leaf characteristics. Then, these characteristics were transferred and a target
to
domain. First, they used the fine-tuning Inception V3 network in the source domain to
learn general plant leaf characteristics. Then, these characteristics were transferred to the
target domain to learn new leaf types from few images. For the FSL method, the DL Sia-
mese network with Triplet loss was utilized. The results demonstrated that dataset size
could be reduced by 89.1% with only a 4% loss in accuracy, that is, this method is good
Agriculture 2021, 11, 707 10 of 18

the target domain to learn new leaf types from few images. For the FSL method, the DL
Siamese network with Triplet loss was utilized. The results demonstrated that dataset size
could be reduced by 89.1% with only a 4% loss in accuracy, that is, this method is good for
small training sets.
The concept of citizen science was proposed in 1995. In this method, nonprofessional
volunteers collect and/or process data as part of a scientific inquiry. In the case of plant
disease and pest classification, farmers and field workers upload the collected images to a
server; then, those images would be properly labeled and processed by an expert [86]. This
idea has been applied in practice. PEAT (a company in Berlin) has built an Android APP
called Plantix that supports farmers with small networks.
Another method for expanding datasets is data sharing. Now, many studies focus
on automatic disease classification around the world. If the various datasets are shared
and properly integrated, then the database will be more representative. This condition will
promote more meaningful and satisfactory research results.

Table 4. Studies on using data augmentation technologies to expand the dataset.

Expanded Dataset Methods Best Accuracy Reference

Direction disturbance and light
From 1053 to 13,689 images disturbance and PCA (Principal 97.62% Bin et al. (2017) [83]
components analysis) jittering
Noise addition, color jittering, and
From 10,820 to 32,460 images 96.17% (improved 3.15%) Lin et al. (2018) [58]
radial blur
From 54,309 to 87,848 images Cropping, resizing 99.53% Ferentinos (2018) [11]
From 1567 to 46,409 images Segmentation, resizing 94.00% (improved 12%) Arnal Barbedo (2019) [86]
From 5000 to 43,398 images Resizing, crop, rotation, noise... 85.98% Fuente et al. (2017) [72]
Affine transformation, perspective
From 4483 to 33,469 images 96.30% Srdjan et al. (2016) [87]
transformation, and rotation

3.2. Nonideal Robustness

In classic DL problems, we often assume that the training and test sets have the same
distribution. Usually, we train the model on the training set and test the model on the
test set. However, the test scenario is often uncontrollable in actual application. The
distribution of the test set is really different from the training set due to various factors,
such as the influence of season and climate. Under the circumstances, the overfitting
problem appears, that is, the trained model does not work well in practical application.
This nonideal robustness problem was confirmed by Mohanty et al. [53], who trained and
tested deep CNN (DCNN) models with the PlantVillage dataset; the top accuracy they
obtained was 99.35%. However, when the DCNN models were tested on a set of images
taken under conditions that were different from the training set, the accuracy dropped to
31% [53]. Similarly, Ferentinos used CNN models (i.e., AlexNet, GoogLeNet, and VGG) to
detect and recognize plant diseases with a public dataset PlantVillage. When the model
was trained and tested with PlantVillage, the best success was 99.53% with the VGG model.
However, when they trained the VGG model with laboratory images and tested it with
field images, the success rate was only up to 33.27% [11].
Three ways can be used to improve the robustness of CNN models. Compressed
models that have a simpler set of parameters show more robustness and less overfit-
ting. However, compressed models achieve poor performance in dealing with complex
recognition. Unsupervised-based DL methods are also good at achieving more robust
performances. Compared with the overall performance of supervised DL models, that
of unsupervised models often drops largely. Another method is multicondition training
(MCT). Yuwana et al. proposed MCT to train more robust DCNNs. They investigated
two types of distortion: blurring and rotations. They evaluated the model on a tea disease
dataset with 5632 images. The results showed that MCT improved the robustness of DCNN
to some extent [59]. Still, another method is persistently enriching the diversity of datasets,
Agriculture 2021, 11, 707 11 of 18

for example through using different geographical locations and cultivation conditions. It is
not a simple task, and social work and cooperation are particularly important.

3.3. Symptom Variations

When detecting plant diseases, we usually assume that the symptoms of the disease
will not change. The symptoms of plant diseases are the results of the interaction of
diseases, plants, and the environment [88]. Changes in any one of the three may lead to
changes in disease symptoms, as discussed below.
In general, plant disease has the following three variations: (1) at different devel-
opment stages of the disease, the symptoms shown may be different [73,88]; (2) in the
same period, multiple diseases may be observed on the same plant leaves. If multiple
diseases are clustered together, then the symptoms may change drastically, which brings
difficulty in identifying the types of diseases [88]; (3) similar symptoms may appear among
different diseases, which increases the difficulty of disease classification. Meanwhile, the
age [89], genotype [90], and healthy tissue color variation (and consequent contrast alter-
ations) [88,91] of the plant itself may cause difficulty in recognizing plant diseases. Other
factors, such as temperature, humidity, wind, soil condition, and sunlight, may also alter
the symptoms of a specific disease.
The interaction of diseases, plants, and the environment may lead to all kinds of
symptom variations, which bring great challenges to image capture and annotation. Two
methods can be used to solve this problem:
1. collecting images of specific diseases that contain the entire range of variation [88]; and
2. gradually enriching the diversity of the database in practical applications [73].
The first method is unrealistic because collecting images of the entire range of variation
is a very labor-intensive and financially demanding task, and whether researchers have
collected variations completely is unclear. The other method is much more realistic, and
this method is currently extensively used by researchers to effectively increase the diversity
of data.

3.4. Image Background

The influence of the picture background on the final classification is unclear. Two
situations should be considered. One is that a regularization process is used when collecting
images, which generates relatively homogeneous backgrounds. In this case, the background
is usually retained. It will not reduce the classification effect and may also improve
the classification accuracy. Mohanty et al. used three different versions of the whole
PlantVillage dataset (color, grayscaled, and segmented) to identify plant diseases and
assess the influence of image background on classification results. The results showed that
the performance of the DCNN model using colored images was slightly higher than that of
the model using the segmented version of the images [53]. The other situation occurs when
images are collected in real-time conditions with a busy background, and some features
of the background are similar to the region of interest. Under these circumstances, leaf
segmentation technology is needed. Otherwise, the model will also learn the features of
the background during training, which will lead to erroneous classification results.
In general, there are five methods that can be used for leaf segmentation. The threshold
segmentation technique, which segments the foreground by setting a specific threshold,
has a serious disadvantage. Usually, the same threshold is used for all pixels, which may
produce incorrect holes or even divide the object into several pieces. This disadvantage will
lead to the subsequent process, such as image classification, being harmed [92]. Meanwhile,
obtaining a reasonable threshold, which is usually selected by manual work, is difficult. K-
means clustering is automatic and works well in most circumstances but is time consuming
and unsuitable for high-speed scenes [93]. Otsu, which is an effective and adaptive
thresholding method, has been widely used for image segmentation [94]. Although the
Otsu method works well with regard to time consumption and is threshold adaptive, it
will not produce an appropriate threshold when the gray-level histogram approximates
Agriculture 2021, 11, 707 12 of 18

a unimodal distribution [95]. One more method is DL FCN. FCN is trained pixel to
pixel on semantic segmentation to achieve the pixel-level classification of images. If we
ignore time and memory limitations, then the FCN method can segment images of any
size but has some drawbacks, such as inadequately considering the relationship between
pixels [96]. The final segmentation method is watershed segmentation, which is an effective
segmentation method. The main drawback of this algorithm is the over-segmentation;
three optimized watershed algorithms, namely, hierarchical watershed segmentation, post-
merging watershed segmentation, and marker-based watershed segmentation [89], have
been proposed to solve this problem. No single segmentation method is suitable for all
problems. The combined use of different methods would be a good choice. Gao and
Lin proposed a fully automatic segmentation method for medicinal plant leaf images in
a complex background. First, they used a vein enhancement and extraction operation
to obtain an accurate foreground marker image. Then, the marker-controlled watershed
method was used to realize image segmentation. The results of the test experiment showed
that the proposed method was better than many other automatic image segmentation
methods, such as DL FCV [96].

4. Discussion
Table 5 provides and explains all the necessary information to help readers choose
one or more criteria and compare different DL models at a glance. As shown in Table 5,
most authors use similar network architectures and thus attain similar experiment results.
Accordingly, new tests with more challenging datasets and new leaner DL architectures
should be implemented; otherwise, much repetition work will appear.

Table 5. Studies on different CNN methods applied to plant leaf disease identification.

No. Reference Task Dataset Method Accuracy Pros and Cons

Mohanty et al. Identify 14 crop species 54,306 images Not good for
1 AlexNet, GoogLeNet 99.35%
(2016) [53] and 26 diseases from PlantVillage practical application
5000 images
Detect diseases and pests
taken under VGGNet and Lacking number of samples, the
Fuentes et al. in tomato plants using 83%
2 different Residual precision would be lower in
(2017) [72] images captured in-place (mean)
conditions Network (ResNet) practical application
by camera devices
and scenarios
Future works will focus on
500 images of rice
Chen et al. Identify rice and maize deploying the module on mobile
3 and 466 images VGGNet, Inception 92%
(2020) [60] leaf diseases devices and applying it to more
of maize
real-world applications
Identify four common The image generation technique
A novel architecture
types of apple leaf 13,689 images of proposed in this paper can
Bin et al. based on AlexNet,
4 diseases (mosaic, rust, diseased 97.62% enhance the robustness of the
(2017) [83] image
brown spot, and Alternaria apple leaves convolutional neural
generation technique
leaf spot) network model
Brahimi et al. Classify nine diseases of
5 14,828 images AlexNet, GoogLeNet 99.18% Lacking number of samples
(2017) [67] tomato leaves
Some of
Recognize two corn leaf DCNN (Deep Only two corn diseases are
Mishra et al. PlantVillage
6 disease (rust, northern Convolutional Neural 88.46% identified and classified, and the
(2020) [55] dataset and some
leaf blight) Network) dataset is not enough.
real-time images
Darwish et al. Diagnose three maize 15,408 images The diversity of dataset is
7 VGG16&19 98.2%
(2019) [56] plant diseases from Kaggle not enough
Ferentinos Plant disease detection 87,848 images AlexNetOWTBn, 99.53% Obtaining significantly high
8
(2018) [11] and diagnosis (PlantVillage) VGG (best) success rate
Train more robust deep Multicondition Only two segmentation methods
Yuwana et al.
9 convolutional 5632 images of tea training (MCT), None- (blur with kernel size of 5 and
(2019) [59]
neural networks AlexNet, GoogLeNet rotation of 40◦ ) were used
Agriculture 2021, 11, 707 13 of 18

As for the unique challenge, insufficient datasets or tedious labeling work, besides the
methods discussed in Section 3.1, unsupervised and semi-supervised model methods may
be a good choice. In the unsupervised models, such as generative adversarial networks
(GANs) [97] and variational autoencoders (VAEs) [98], only normal samples are used
for training, which solves the problem of difficulty in obtaining disease datasets. The
existing few-shot classification studies are mainly based on supervised learning schemes,
ignoring the helpful information of unlabeled samples [99]. However, the semi-supervised
algorithms use both a few annotated samples and many unannotated samples to train a
model and can use unlabeled samples to solve the difficulty of network training in the case
of a few labeled samples. Therefore, the use of unsupervised and semi-supervised model
methods may be a good research direction in the future.
As for the network design, the models proposed between 2017 and 2021 are slightly
different from the earlier ones. They are specially focused on reducing the number of
networks parameters [94], designing the networks to be trained with a small database [88],
and designing the networks to be trained with field images [100]. Undoubtedly, the trend
of designing computationally efficient classification networks will continue to develop in
the future [101].
Today, the quick development of intelligent devices, such as smartphones, personal
computers, fixed cameras, and UAV, is making image classification projects more conve-
nient and intelligent. He et al. proposed a scheme based on the combination of android
clients and servers, which are ubiquitous in our daily lives. The scheme consists of two
parts: (1) mobile phone client, through which users can upload the collected images to the
server; (2) server-side program, which processes the images and returns the classification
results to the user. Meanwhile, the server also needs to store the relevant results in the
database to facilitate the query of users [102]. Turui (Beijing, China) Information Technol-
ogy Co., LTD. (https://www.mapsharp.com/wzsy, accessed on 22 July 2021) developed
the “Insect Prophet” pest monitoring product. Using the cloud platform, it can easily
realize the functions of taking photos to identity pests and counting insects. With the
quick development of intelligent devices, the application of deep learning in daily life will
become more and more extensive. However, agricultural areas are sometimes far from
well-connected regions. Under this circumstance, edge devices and mobile clients, which
do not need to send data to the server and can be deployed offline, could be great measures.
Meanwhile, some research shows [103,104] that the electrical signal response pro-
duced within plants can be used for real-time detection of plant diseases. Plants perceive
the environment by generating electrical signals that essentially represent changes in un-
derlying physiological processes [105]. Under the influence of stress (such as disease), the
metabolic activities of various cells and tissues of plants are unstable, which is bound to
be reflected in physiological electrical properties. Therefore, the extraction of meaningful
features from the generating electrical signals (such as the varying capacitance, conduc-
tivity, impedance) and the use of such extracted features [106] would be a good research
direction for the classification of plant diseases. For example, Najdenovska et al. used plant
electrophysiological signals recorded from 12 tomato plants contaminated with spider
mites for an automated classification of the plant’s abnormal state caused by spider mites,
and this study got an accuracy of 80% [104].

5. Conclusions
DL methods have gained widespread application in plant disease detection and
classification. It has solved or partially solved the problems of traditional machine learning
methods. DL, which is a branch of machine learning, is mainly used for image classification,
target detection, and image segmentation. In this paper, we reviewed the latest CNN
networks pertinent to plant leaf disease classification. We introduce the process of CNN
methods applied to plant disease classification and summarize DL principles involved
in plant disease classification. We also summarize some problems and corresponding
solutions of DL used for plant disease classification with extrinsic and intrinsic factors as
Agriculture 2021, 11, 707 14 of 18

listed below: (1) insufficient datasets: transfer learning, data augmentation techniques,
citizen science, and data sharing; (2) no-ideal robustness: compressed model, unsupervised
DL model, and multicondition training; (3) symptom variations: collecting an entire
range of variation and gradually enriching the diversity of dataset; (4) image background:
threshold segmentation technique, K-means clustering, Otsu, DL FCN, and watershed
segmentation. Furthermore, we discussed the future development direction in plant disease
classification, for example, plant electrophysiology and the combination of the mobile
phone client and the server-side program would be good future research directions [106].
Such a combination is good for the practice and real-time application of DL methods in
plant disease classification.

Author Contributions: Conceptualization, J.L. and H.J.; supervision, J.L.; visualization, J.L. and H.J.;
writing—original draft preparation, L.T.; writing—review and editing, L.T. and J.L. All authors have
read and agreed to the published version of the manuscript.
Funding: This research was funded by the Sichuan Science and Technology Program under grant
2021YFN0020; the Innovation Fund of Postgraduate, Xihua University under grant YCJJ2020041;
the Sichuan Science and Technology Program under grant 2019YFN0106; the Key Project of Xihua
University under grant DC1900007141, and the National Natural Science Foundation of China under
grant 31870347.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Mukti, I.Z.; Biswas, D. Transfer Learning Based Plant Diseases Detection Using ResNet50. In Proceedings of the 2019 4th Interna-
tional Conference on Electrical Information and Communication Technology (EICT), Khulna, Bangladesh, 20–22 December 2019.
2. Arunnehru, J.; Vidhyasagar, B.S.; Anwar Basha, H. Plant Leaf Diseases Recognition Using Convolutional Neural Network and
Transfer Learning. In International Conference on Communication, Computing and Electronics Systems; Bindhu, V., Chen, J., Tavares,
J.M.R.S., Eds.; Springer: Singapore, 2020; pp. 221–229.
3. Hughes, D.P.; Salathe, M. An open access repository of images on plant health to enable the development of mobile disease
diagnostics. arXiv 2015, arXiv:1511.08060.
4. Kianat, J.; Khan, M.A.; Sharif, M.; Akram, T.; Rehman, A.; Saba, T. A joint framework of feature reduction and robust feature
selection for cucumber leaf diseases recognition. Optik 2021, 240, 166566. [CrossRef]
5. Zhang, S.; Zhang, S.; Zhang, C.; Wang, X.; Shi, Y. Cucumber leaf disease identification with global pooling dilated convolutional
neural network. Comput. Electron. Agric. 2019, 162, 422–430. [CrossRef]
6. Agarwal, M.; Gupta, S.; Biswas, K.K. A new conv2d model with modified relu activation function for identification of disease
type and severity in cucumber plant. Sustain. Comput. Inform. Syst. 2021, 30, 100473.
7. Chen, J.; Zhang, D.; Zeb, A.; Nanehkaran, Y.A. Identification of rice plant diseases using lightweight attention networks. Expert
Syst. Appl. 2021, 169, 114514. [CrossRef]
8. Shrivastava, V.K.; Pradhan, M.K.; Minz, S.; Thakur, M.P. Rice plant disease classification using transfer learning of deep
convolution neural network. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2019, XLII-3/W6, 631–635. [CrossRef]
9. Sun, H.; Zhai, L.; Teng, F.; Li, Z.; Zhang, Z. qRgls1. 06, a major QTL conferring resistance to gray leaf spot disease in maize. Crop.
J. 2021, 9, 342–350. [CrossRef]
10. Yu, C.; Ai-hong, Z.; Ai-Jun, R.; Hong-qin, M. Types of Maize virus diseases and progress in virus identification techniques in
China. J. Northeast Agric. Univ. 2014, 21, 75–83. [CrossRef]
11. Ferentinos, K. Deep learning models for plant disease detection and diagnosis. Comput. Electron. Agric. 2018, 145, 311–318.
[CrossRef]
12. Abbas, A.; Jain, S.; Gour, M.; Vankudothu, S. Tomato plant disease detection using transfer learning with C-GAN synthetic
images. Comput. Electron. Agric. 2021, 187, 106279. [CrossRef]
13. Sankaran, S.; Mishra, A.; Ehsani, R.; Davis, C. A review of advanced techniques for detecting plant diseases. Comput. Electron.
Agric. 2010, 72, 1–13. [CrossRef]
14. Barbedo Arnal, J.G. An Automatic Method to Detect and Measure Leaf Disease Symptoms Using Digital Image Processing. Plant
Dis. 2014, 98, 1709–1716. [CrossRef]
Agriculture 2021, 11, 707 15 of 18

15. Feng, Q.; Dongxia, L.; Bingda, S.; Liu, R.; Zhanhong, M.; Haiguang, W. Identification of Alfalfa Leaf Diseases Using Image
Recognition Technology. PLoS ONE 2016, 11, e168274.
16. Omrani, E.; Khoshnevisan, B.; Shamshirband, S.; Saboohi, H.; Anuar, N.B.; Nasir, M.H.N.M. Potential of radial basis function-
based support vector regression for apple disease detection. Measurement 2014, 55, 512–519. [CrossRef]
17. Barbedo Arnal, J.G. A new automatic method for disease symptom segmentation in digital photographs of plant leaves. Eur. J.
Plant Pathol. 2017, 147, 349–364. [CrossRef]
18. Springer. SVM-Based Detection of Tomato Leaves Diseases; Springer: Berlin/Heidelberg, Germany, 2015.
19. Rumpf, T.; Mahlein, A.K.; Steiner, U.; Oerke, E.C.; Dehne, H.W.; Plümer, L. Early detection and classification of plant diseases
with Support Vector Machines based on hyperspectral reflectance. Comput. Electron. Agric. 2010, 74, 91–99. [CrossRef]
20. Hossain, E.; Hossain, M.F.; Rahaman, M.A. A Color and Texture Based Approach for the Detection and Classification of Plant Leaf
Disease Using KNN Classifier. In Proceedings of the 2019 International Conference on Electrical, Computer and Communication
Engineering (ECCE), Cox’s Bazar, Bangladesh, 7–9 February 2019.
21. Mohana, R.M.; Reddy CK, K.; Anisha, P.R.; Murthy, B.R. Random forest algorithms for the classification of tree-based ensemble.
Mater. Today Proc. 2021. [CrossRef]
22. Türkoğlu, M.; Hanbay, D. Plant disease and pest detection using deep learning-based features. Turk. J. Electr. Eng. Comput. Sci.
2019, 27, 1636–1651. [CrossRef]
23. Arivazhagan, S.; Shebiah, R.N.; Ananthi, S.; Varthini, S.V. Detection of unhealthy region of plant leaves and classification of plant
leaf diseases using texture features. Agric. Eng. Int. CIGR J. 2013, 15, 211–217.
24. Jiang, F.; Lu, Y.; Chen, Y.; Cai, D.; Li, G. Image recognition of four rice leaf diseases based on deep learning and support vector
machine. Comput. Electron. Agric. 2020, 179, 105824. [CrossRef]
25. Gao, J.; French, A.P.; Pound, M.P.; He, Y.; Pridmore, T.P.; Pieters, J.G. Deep convolutional neural networks for image-based
Convolvulus sepium detection in sugar beet fields. Plant Methods 2020, 16, 29. [CrossRef] [PubMed]
26. Athanikar, G.; Badar, P. Potato leaf diseases detection and classification system. Int. J. Comput. Sci. Mob. Comput. 2016, 5, 76–88.
27. Yu, J.; Sharpe, S.M.; Schumann, A.W.; Boyd, N.S. Deep learning for image-based weed detection in turfgrass. Eur. J. Agron. 2019,
104, 78–84. [CrossRef]
28. Bansal, P.; Kumar, R.; Kumar, S. Disease Detection in Apple Leaves Using Deep Convolutional Neural Network. Agriculture 2021,
11, 617. [CrossRef]
29. Wang, H.; Raj, B. On the origin of deep learning. arXiv 2017, arXiv:1702.07800.
30. Deng, L.; Yu, D. Deep Learning: Methods and Applications. Found. Trends Signal Process. 2014, 7, 197–387. [CrossRef]
31. Dyrmann, M.; Karstoft, H.; Midtiby, H.S. Plant species classification using deep convolutional neural network. Biosyst. Eng. 2016,
151, 72–80. [CrossRef]
32. Kussul, N.; Lavreniuk, M.; Skakun, S.; Shelestov, A. Deep Learning Classification of Land Cover and Crop Types Using Remote
Sensing Data. IEEE Geosci. Remote. Sens. Lett. 2017, 14, 778–782. [CrossRef]
33. Wason, R. Deep learning: Evolution and expansion. Cogn. Syst. Res. 2018, 52, 701–708. [CrossRef]
34. Geetharamani, G.; Pandian, A. Identification of plant leaf diseases using a nine-layer deep convolutional neural network. Comput.
Electr. Eng. 2019, 76, 323–338.
35. Huang, G.; Liu, Z.; Laurens, V.; Weinberger, K.Q. Densely connected convolutional networks. In Proceedings of the IEEE
Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 4700–4708. Available
online: https://openaccess.thecvf.com/content_cvpr_2017/papers/Huang_Densely_Connected_Convolutional_CVPR_2017
_paper.pdf (accessed on 22 July 2021).
36. Too, E.C.; Yujian, L.; Njuki, S.; Yingchun, L. A comparative study of fine-tuning deep learning models for plant disease
identification. Comput. Electron. Agric. 2018, 161, 272–279. [CrossRef]
37. He, K.; Zhang, X.; Ren, S.; Jian, S. Identity Mappings in Deep Residual Networks. In European Conference on Computer Vision;
Springer: Cham, Switzerland, 2016.
38. Guan, S.; Kamona, N.; Loew, M. Segmentation of Thermal Breast Images Using Convolutional and Deconvolutional Neural
Networks. In Proceedings of the 2018 IEEE Applied Imagery Pattern Recognition Workshop (AIPR), Washington, DC, USA,
9–11 October 2018.
39. Fakhry, A.; Zeng, T.; Ji, S. Residual Deconvolutional Networks for Brain Electron Microscopy Image Segmentation. IEEE Trans.
Med. Imaging 2017, 36, 447–456. [CrossRef] [PubMed]
40. Liu, J.; Wang, Y.; Li, Y.; Fu, J.; Li, J.; Lu, H. Collaborative Deconvolutional Neural Networks for Joint Depth Estimation and
Semantic Segmentation. IEEE Trans. Neural Netw. Learn. Syst. 2018, 29, 5655–5666. [CrossRef] [PubMed]
41. Wang, J.; Wang, Z.; Tao, D.; See, S.; Wang, G. Learning Common and Specific Features for RGB-D Semantic Segmentation with
Deconvolutional Networks. In European Conference on Computer Vision; Springer: Cham, Switzerland, 2016.
42. Gehlot, S.; Gupta, A.; Gupta, R. SDCT-AuxNet: DCT Augmented Stain Deconvolutional CNN with Auxiliary Classifier for
Cancer Diagnosis. Med. Image Anal. 2020, 61, 101661. [CrossRef]
43. Duggal, R.; Gupta, A.; Gupta, R.; Mallick, P. SD-Layer: Stain Deconvolutional Layer for CNNs in Medical Microscopic Imaging;
Springer: Cham, Switzerland, 2017.
44. Gu, J.; Wang, Z.; Kuen, J.; Ma, L.; Wang, G. Recent Advances in Convolutional Neural Networks. Pattern Recognit. 2015, 77,
354–377. [CrossRef]
Agriculture 2021, 11, 707 16 of 18

45. Krizhevsky, A.; Sutskever, I.; Hinton, G. ImageNet Classification with Deep Convolutional Neural Networks. Adv. Neural Inf.
Process. Syst. 2012, 25, 1097–1105. [CrossRef]
46. Gao, Z.; Luo, Z.; Zhang, W.; Lv, Z.; Xu, Y. Deep Learning Application in Plant Stress Imaging: A Review. AgriEngineering 2020, 2,
430–446. [CrossRef]
47. Bengio, Y.; Simard, P.; Frasconi, P. Learning long-term dependencies with gradient descent is difficult. IEEE Trans. Neural Netw.
1994, 5, 157–166. [CrossRef] [PubMed]
48. Glorot, X.; Bengio, Y. Understanding the difficulty of training deep feedforward neural networks. J. Mach. Learn. Res. 2010, 9,
249–256.
49. He, K.; Zhang, X.; Ren, S.; Sun, J. Deep Residual Learning for Image Recognition. arXiv 2016, arXiv:1512.03385.
50. Howard, A.G.; Zhu, M.; Chen, B.; Kalenichenko, D.; Wang, W.; Weyand, T.; Andreetto, M.; Adam, H. MobileNets: Efficient
Convolutional Neural Networks for Mobile Vision Applications. arXiv 2017, arXiv:1704.04861.
51. Tan, M.; Le, Q.V. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. arXiv 2019, arXiv:1704.04861.
52. Parvat, A.; Chavan, J.; Kadam, S.; Dev, S.; Pathak, V. A survey of deep-learning frameworks: 2017 International Conference on
Inventive Systems and Control (ICISC). In Proceedings of the 2017 International Conference on Inventive Systems and Control
(ICISC), Coimbatore, India, 19–20 January 2017.
53. Mohanty, S.P.; Hughes, D.P.; Marcel, S. Using Deep Learning for Image-Based Plant Disease Detection. Front. Plant Sci. 2016,
7, 1419. [CrossRef]
54. Brahimi, M.; Mahmoudi, S.; Boukhalfa, K.; Moussaoui, A. Deep interpretable architecture for plant diseases classification. In
Proceedings of the 2019 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA), Poznan, Poland,
18–20 September 2019.
55. Mishra, S.; Sachan, R.; Rajpal, D. Deep Convolutional Neural Network based Detection System for Real-time Corn Plant Disease
Recognition. Procedia Comput. Ence 2020, 167, 2003–2010. [CrossRef]
56. Darwish, A.; Ezzat, D.; Hassanien, A.E. An optimized model based on convolutional neural networks and orthogonal learning
particle swarm optimization algorithm for plant diseases diagnosis. Swarm Evol. Comput. 2019, 52, 100616. [CrossRef]
57. Amanda, R.; Kelsee, B.; Peter, M.C.; Babuali, A.; James, L.; Hughes, D.P. Deep Learning for Image-Based Cassava Disease
Detection. Front. Plant Sci. 2017, 8, 1852.
58. Lin, Z.; Mu, S.; Shi, A.; Pang, C.; Student, G.; Sun, X.; Student, G. A Novel Method of Maize Leaf Disease Image Identification
Based on a Multichannel Convolutional Neural Network. Trans. ASABE 2018, 61, 1461–1474. [CrossRef]
59. Yuwana, R.S.; Suryawati, E.; Zilvan, V.; Ramdan, A.; Fauziah, F. Multi-Condition Training on Deep Convolutional Neural
Networks for Robust Plant Diseases Detection. In Proceedings of the 2019 International Conference on Computer, Control,
Informatics and its Applications (IC3INA), Tangerang, Indonesia, 23–24 October 2019.
60. Chen, J.; Chen, J.; Zhang, D.; Sun, Y.; Nanehkaran, Y.A. Using deep transfer learning for image-based plant disease identification.
Comput. Electron. Agric. 2020, 173, 105393. [CrossRef]
61. Fujita, E.; Kawasaki, Y.; Uga, H.; Kagiwada, S.; Iyatomi, H. Basic investigation on a robust and practical plant diagnostic system.
In Proceedings of the 2016 15th IEEE International Conference on Machine Learning and Applications (ICMLA), Anaheim, CA,
USA, 18–20 December 2016.
62. Ghazi, M.M.; Yanikoglu, B.; Aptoula, E. Plant identification using deep neural networks via optimization of transfer learning
parameters. Neurocomputing 2017, 235, 228–235. [CrossRef]
63. Hidayatuloh, A.; Nursalman, M.; Nugraha, E. Identification of Tomato Plant Diseases by Leaf Image Using Squeezenet Model.
In Proceedings of the 2018 International Conference on Information Technology Systems and Innovation (ICITSI), Bandung,
Indonesia, 22–26 October 2018.
64. Juncheng, M.; Keming, D.; Feixiang, Z.; Lingxian, Z.; Zhihong, G.; Zhongfu, S. A recognition method for cucumber diseases using
leaf symptom images based on deep convolutional neural network. Comput. Electron. Agric. 2018, 154, 18–24.
65. Bollis, E.; Pedrini, H.; Avila, S. Weakly Supervised Learning Guided by Activation Mapping Applied to a Novel Citrus Pest
Benchmark. In Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW),
Seattle, WA, USA, 14–19 June 2020.
66. Ge, M.; Su, F.; Zhao, Z.; Su, D. Deep Learning Analysis on Microscopic Imaging in Materials Science. Mater. Today Nano 2020,
11, 100087. [CrossRef]
67. Brahimi, M.; Boukhalfa, K.; Moussaoui, A. Deep Learning for Tomato Diseases: Classification and Symptoms Visualization. Appl.
Artif. Intell. 2017, 31, 1–17. [CrossRef]
68. Arsenovic, M.; Karanovic, M.; Sladojevic, S.; Anderla, A.; Stefanovic, D. Solving Current Limitations of Deep Learning Based
Approaches for Plant Disease Detection. Symmetry 2019, 11, 939. [CrossRef]
69. Peng, Y.; Wang, Y. An industrial-grade solution for agricultural image classification tasks. Comput. Electron. Agric. 2021,
187, 106253. [CrossRef]
70. Wang, Y.; Wang, J.; Zhang, W.; Zhan, Y.; Guo, S.; Zheng, Q.; Wang, X. A survey on deploying mobile deep learning applications:
A systemic and technical perspective. Digit. Commun. Netw. 2021. [CrossRef]
71. Gao, J.; Westergaard, J.C.; Sundmark, E.H.R.; Bagge, M.; Liljeroth, E.; Alexandersson, E. Automatic late blight lesion recognition
and severity quantification based on field imagery of diverse potato genotypes by deep learning. Knowl. Based Syst. 2021,
214, 106723. [CrossRef]
Agriculture 2021, 11, 707 17 of 18

72. Fuentes, A.; Yoon, S.; Kim, S.C.; Park, D.S. A Robust Deep-Learning-Based Detector for Real-Time Tomato Plant Diseases and
Pests Recognition. Sensors 2017, 17, 2022. [CrossRef]
73. Barbedo, J.G.A. Factors influencing the use of deep learning for plant disease recognition. Biosyst. Eng. 2018, 172, 84–91.
[CrossRef]
74. Coulibaly, S.; Kamsu-Foguem, B.; Kamissoko, D.; Traore, D. Deep neural networks with transfer learning in millet crop images.
Comput. Ind. 2019, 108, 115–120. [CrossRef]
75. Abdalla, A.; Cen, H.; Wan, L.; Rashid, R.; He, Y. Fine-tuning convolutional neural network with transfer learning for semantic
segmentation of ground-level oilseed rape images in a field with high weed pressure. Comput. Electron. Agric. 2019, 167, 105091.
[CrossRef]
76. Thenmozhi, K.; Reddy, U.S. Crop pest classification based on deep convolutional neural network and transfer learning. Comput.
Electron. Agric. 2019, 164, 104906. [CrossRef]
77. Suh, H.K.; IJsselmuiden, J.; Hofstee, J.W.; van Henten, E.J. Transfer learning for the classification of sugar beet and volunteer
potato under field conditions. Biosyst. Eng. 2018, 174, 50–65. [CrossRef]
78. Hendrycks, D.; Mu, N.; Cubuk, E.D.; Zoph, B.; Gilmer, J.; Lakshminarayanan, B. AugMix: A Simple Data Processing Method to
Improve Robustness and Uncertainty. arXiv 2019, arXiv:1912.02781.
79. Ho, D.; Liang, E.; Stoica, I.; Abbeel, P.; Chen, X. Population Based Augmentation: Efficient Learning of Augmentation Policy
Schedules. In Proceedings of the International Conference on Machine Learning, Long Beach, CA, USA, 9–15 June 2019;
pp. 2731–2741.
80. Lim, S.; Kim, I.; Kim, T.; Kim, C.; Kim, S. Fast AutoAugment. Adv. Neural Inf. Process. Syst. 2019, 32, 6665–6675.
81. Cubuk, E.D.; Zoph, B.; Shlens, J.; Le, Q.V. Randaugment: Practical automated data augmentation with a reduced search space. In
Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Seattle, WA,
USA, 14–19 June 2020.
82. Yun, S.; Han, D.; Chun, S.; Oh, S.J.; Yoo, Y.; Choe, J. CutMix: Regularization Strategy to Train Strong Classifiers with Localizable
Features. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Seoul, Korea, 27–28 October 2019;
pp. 6023–6032. Available online: https://openaccess.thecvf.com/content_ICCV_2019/papers/Yun_CutMix_Regularization_
Strategy_to_Train_Strong_Classifiers_With_Localizable_Features_ICCV_2019_paper.pdf (accessed on 22 July 2021).
83. Bin, L.; Yun, Z.; Dongjian, H.; Yuxiang, L. Identification of Apple Leaf Diseases Based on Deep Convolutional Neural Networks.
Symmetry 2017, 10, 11.
84. Douarre, C.; Crispim-Junior, C.F.; Gelibert, A.; Tougne, L.; Rousseau, D. Novel data augmentation strategies to boost supervised
segmentation of plant disease. Comput. Electron. Agric. 2019, 165, 104967. [CrossRef]
85. Argüeso, D.; Picon, A.; Irusta, U.; Medela, A.; Alvarez-Gila, A. Few-Shot Learning approach for plant disease classification using
images taken in the field. Comput. Electron. Agric. 2020, 175, 105542. [CrossRef]
86. Arnal Barbedo, J.G. Plant disease identification from individual lesions and spots using deep learning. Biosyst. Eng. 2019, 180,
96–107. [CrossRef]
87. Srdjan, S.; Marko, A.; Andras, A.; Dubravko, C.; Darko, S. Deep Neural Networks Based Recognition of Plant Diseases by Leaf
Image Classification. Comput. Intell. Neurosci. 2016, 2016, 3289801.
88. Barbedo Arnal, J.G. A review on the main challenges in automatic plant disease identification based on visible range images.
Biosyst. Eng. 2016, 144, 52–60. [CrossRef]
89. Zhang, H.; Tang, Z.; Xie, Y.; Gao, X.; Chen, Q. A watershed segmentation algorithm based on an optimal marker for bubble size
measurement. Measurement 2019, 138, 182–193. [CrossRef]
90. Peressotti, E.; Duchene, E.; Merdinoglu, D.; Mestre, P. A semi-automatic non-destructive method to quantify grapevine downy
mildew sporulation. J. Microbiol. Methods 2011, 84, 265–271. [CrossRef] [PubMed]
91. Clément, A.; Verfaille, T.; Lormel, C.; Jaloux, B. A new colour vision system to quantify automatically foliar discolouration caused
by insect pests feeding on leaf cells. Biosyst. Eng. 2015, 133, 128–140. [CrossRef]
92. Ling, Q.; Yan, J.; Li, F.; Zhang, Y. A background modeling and foreground segmentation approach based on the feedback of
moving objects in traffic surveillance systems. Neurocomputing 2014, 133, 32–45. [CrossRef]
93. Du, H.; Chen, X.; Xi, J. An improved background segmentation algorithm for fringe projection profilometry based on Otsu
method. Opt. Commun. 2019, 453, 124206. [CrossRef]
94. Singh, B.; Toshniwal, D.; Allur, S.K. Shunt connection: An intelligent skipping of contiguous blocks for optimizing MobileNet-V2.
Neural Netw. 2019, 118, 192–203. [CrossRef]
95. Xu, X.; Xu, S.; Jin, L.; Song, E. Characteristic analysis of Otsu threshold and its applications. Pattern Recognit. Lett. 2011, 32,
956–961. [CrossRef]
96. Gao, L.; Lin, X. Fully automatic segmentation method for medicinal plant leaf images in complex background. Comput. Electron.
Agric. 2019, 164, 104924. [CrossRef]
97. Goodfellow, I.J.; Pouget-Abadie, J.; Mirza, M.; Xu, B.; Warde-Farley, D.; Ozair, S.; Courville, A.; Bengio, Y. Generative Adversarial
Networks. Adv. Neural Inf. Process. Syst. 2014, 3, 2672–2680.
98. Gunduz, H. An efficient dimensionality reduction method using filter-based feature selection and variational autoencoders on
Parkinson’s disease classification. Biomed. Signal Process. Control. 2021, 66, 102452. [CrossRef]
Agriculture 2021, 11, 707 18 of 18

99. Li, Y.; Chao, X. Semi-supervised few-shot learning approach for plant diseases recognition. Plant Methods 2021, 17, 68. [CrossRef]
[PubMed]
100. Anami, B.S.; Malvade, N.N.; Palaiah, S. Deep learning approach for recognition and classification of yield affecting paddy crop
stresses using field images. Artif. Intell. Agric. 2020, 4, 12–20. [CrossRef]
101. Rahman, S.; Wang, L.; Sun, C.; Zhou, L. Deep Learning Based HEp-2 Image Classification: A Comprehensive Review. Med. Image
Anal. 2020, 65, 101764. [CrossRef] [PubMed]
102. He, Y.; Zhou, Z.; Tian, L.; Liu, Y.; Luo, X. Brown rice planthopper (Nilaparvata lugens Stal) detection based on deep learning.
Precis. Agric. 2020, 21, 1385–1402. [CrossRef]
103. Shre, K.C. An Approach towards Plant Electrical Signal Based External Stimuli Monitoring System. Ph.D. Thesis, University of
Southampton, Southampton, UK, 2017.
104. Najdenovska, E.; Dutoit, F.; Tran, D.; Plummer, C.; Raileanu, L.E. Classification of Plant Electrophysiology Signals for Detection of
Spider Mites Infestation in Tomatoes. Appl. Sci. 2021, 11, 1414. [CrossRef]
105. Chatterjee, S.K.; Das, S.; Maharatna, K.; Masi, E.; Santopolo, L.; Mancuso, S.; Vitaletti, A. Exploring strategies for classification of
external stimuli using statistical features of the plant electrical response. J. R. Soc. Interface 2015, 12, 20141225. [CrossRef]
106. Chatterjee, S.K.; Malik, O.; Gupta, S. Chemical Sensing Employing Plant Electrical Signal Response-Classification of Stimuli
Using Curve Fitting Coefficients as Features. Biosensors 2018, 8, 83. [CrossRef] [PubMed]