agronomy
Article
An AI Based Approach for Medicinal Plant Identification Using
Deep CNN Based on Global Average Pooling
Rahim Azadnia 1 , Mohammed Maitham Al-Amidi 2 , Hamed Mohammadi 3 , Mehmet Akif Cifci 4 , Avat Daryab 5
and Eugenio Cavallo 6, *
Citation: Azadnia, R.; Al-Amidi,
M.M.; Mohammadi, H.; Cifci, M.A.;
Daryab, A.; Cavallo, E. An AI Based
Approach for Medicinal Plant
Identification Using Deep CNN
Based on Global Average Pooling.
Agronomy 2022, 12, 2723. https://
doi.org/10.3390/agronomy12112723
Academic Editors: Ahmed Kayad
and Ahmed Rady
Received: 16 September 2022
Accepted: 26 October 2022
Published: 2 November 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Agronomy 2022, 12, 2723. https://doi.org/10.3390/agronomy12112723
1 Department of Biosystems Engineering, University of Tehran, Karaj 3158777871, Iran
2 Information Technology, Al-Mustaqbal University College, Babylon 51001, Iraq
3 Department of Industrial Engineering and Management Systems, University of Central Florida,
Orlando, FL 32816, USA
4 Department of Computer Engineering, Bandirma Onyedi Eylul University, Balikesir 10200, Turkey
5 Department of Agricultural, Karaj Branch, Islamic Azad University, Karaj 3158777871, Iran
6 Institute of Sciences and Technologies for Sustainable Energy and Mobility (STEMS), National Research
Council (CNR) of Italy, 10129 Torino, Italy
* Correspondence: eugenio.cavallo@cnr.it
Abstract: Medicinal plants have always been studied and considered due to their high importance
for preserving human health. However, identifying medicinal plants is very time-consuming, tedious
and requires an experienced specialist. Hence, a vision-based system can support researchers and
ordinary people in recognising herb plants quickly and accurately. Thus, this study proposes an
intelligent vision-based system to identify herb plants by developing an automatic Convolutional
Neural Network (CNN). The proposed Deep Learning (DL) model consists of a CNN block for feature
extraction and a classifier block for classifying the extracted features. The classifier block includes a
Global Average Pooling (GAP) layer, a dense layer, a dropout layer, and a softmax layer. The solution
has been tested on 3 levels of definitions (64 × 64, 128 × 128 and 256 × 256 pixel) of images for leaf
recognition of five different medicinal plants. As a result, the vision-based system achieved more
than 99.3% accuracy for all the image definitions. Hence, the proposed method effectively identifies
medicinal plants in real-time and is capable of replacing traditional methods.
Keywords: medicinal plant; identification; image processing; Global Average Pooling (GAP); Convo-
lutional Neural Network (CNN)
1. Introduction
Medicinal plants have been used in traditional medicine practices for a long time
because of their nutrients and medicinal properties [1]. Due to their bioactive compounds,
such as phenolic, carotenoid, anthocyanin, and other bio-active components, they are
known for their antioxidant, anti-allergic, anti-inflammatory, and antibacterial properties [2].
Different species of plants are recognized as having medical properties; they can be trees,
shrubs or herbs. Their single diffusion depends on the environmental conditions they
adapted to over time. According to the statistics, about 14–28% of all plants have medicinal
uses [3]. Furthermore, about 3–5% of patients in developed countries, and over 80% in the
rural population in developing countries and about 85% of the population in the Southern
Sahara use medicinal plants to treat their diseases because of their properties [4]. Moreover,
in developed countries, some people have turned to traditional medicines prepared from
medicinal plants to treat and control illnesses and diseases after considering the harmfulness
and side effects of chemical drugs [5,6]. In addition to their medicinal uses, these plants can
be used as food, beverages, and even in the form of cosmetics [7,8]. Unfortunately, many
counterfeit, low-quality, damaged, or not perfectly preserved medicinal plants are being
manufactured and distributed worldwide, which may harm their users [9].
https://www.mdpi.com/journal/agronomy
Agronomy 2022, 12, 2723 2 of 16

Botanists have long identified various species of medicinal plants using traditional
and experience-based methods. Nevertheless, visually and manually identifying medicinal
plants from other similar plants can be extremely challenging and time-consuming for
inexperienced people [10–12].
Plants are generally classified based on their various organs, including roots, flowers,
and leaves. The leaves, one of the most important organs of plants, largely differ among
species and varieties in colors, shapes, and texture characteristics. However, in some cases,
there have always been challenges associated with identifying medicinal plants due to the
apparent similarities of their leaves. Furthermore, the leaf color cannot be considered a
suitable option for plant classification due to their similarities and, most importantly, their
variability during the growing period.
Several real-time vision systems have recently been developed using machine vision
applications and computational methods to identify medicinal plants [13–16]. Deep learn-
ing (DL) techniques have largely been applied since they can simultaneously handle feature
extraction and image selection. As a result, DL has gained great popularity in various
agriculture automation applications over the last few years, where object detection and
image classification are required [17–20].
Convolutional Neural Network (CNN) is one of the most successful DL methods in
which it has been associated with excellent efficiency in image segmentation and pattern recog-
nition [21]. The CNN uses expertly trained layers that have been adopted in several studies
for plant identification and classification [22–25] and plant disease identification [26–29]. CNN
consists of a hierarchy of self-learning properties, in which low-level features such as colors,
corners, and edges are learned from the first layer and high-level features such as textures
and objects within the image are learned from the deep layer. Automatic feature learning
has reduced the sensitivity of CNNs to environmental variables such as light changes.
These models combine the learning of extracted features with the classification, which are
both crucial steps in the image processing process. Therefore, unlike conventional machine
learning algorithms, the manual (hand-crafted) extraction of features is not required.
Nasiri et al. [30] showed that the DL algorithm applied to imaging recognition in
the visible area (400–700 nm) could discriminate grape leaves among 6 different cultivars
with 99% accuracy. Paulson and Ravishankar [31] identified 64 types of medicinal plants
using digital images adopting three DL models of CNN, VGG16, and VGG19, achieving
an accuracy rate of 95.7, 97.8, and 97.6%, respectively. Grinblat et al. [32] used a 3-layer
CNN model to identify three different plants based on their leaf vein patterns and attained
an accuracy recognition rate of 92%. Hu et al. [33] developed a Multi-Scale Function
(MSF) MSF-CNN model to classify plant leaves by integrating multi-scale features with
CNNs. MSF-CNNs consist of multiple learning branches with different scales of learning.
They performed experiments using two public datasets of plant leaves, MalayaKew (MK)
and LeafSnap. Their results demonstrated that the proposed MSF-CNN method is more
effective in recognizing plant leaves than state-of-the-art methods.
It is still difficult to distinguish medicinal herbs from other plants, even with vision-
based systems that have significantly improved their ability to extract complex features
and select the most important ones via conventional Machine Learning (ML). Therefore,
the main objective of the current study was to develop a real-time automatic vision system
for identifying medical plants using a proposed DL algorithm coupled with a machine
vision technique. To put it more clearly, this research aimed to improve the automatic
identification of medicinal plants as their popularity is growing and they are increasingly
being used for artisanal and industrial purposes.

2. Materials and Methods

2.1. Sampling Protocol
The leaves of five medicinal plants, Lemon Balm (Melissa officinalis L.), Stevia (Stevia
rebaudiana Bertoni), Peppermint (Mentha balsamea Wild), Bael (Aegle marmelos) and Tulsi
(Ocimum sanctum L) were collected in the Northern area of Iran, in the cities of Salmas
 being used for artisanal and industrial purposes.

2. Materials and Methods

2.1. Sampling Protocol
The leaves of five medicinal plants, Lemon Balm (Melissa officinalis L.), Stevia (Stevia
Agronomy 2022, 12, 2723 3 of 16
rebaudiana Bertoni), Peppermint (Mentha balsamea Wild), Bael (Aegle marmelos) and
Tulsi (Ocimum sanctum L) were collected in the Northern area of Iran, in the cities of
Salmas (38° 11′ 41′′ N and 44° 45′ 53′′ E), Khoy ( 38° 33′ 01′′ N and 44° 57′ 08′′ E), Maku
(38 ◦ 0 41 00 N and ◦ 0 53 00 ′ Khoy ◦ 330 0100 N and 44 ◦ 57′ 0 08′′
00 E), Maku° (39′◦ 170′′44.9600 N
(39°11
17′ 44.96 ′′
N 44
and4544° 30E),47.64′′ E)(38and Urmia (37° 32 55 N and 45 04 03 E). The
and 44 ◦ 30 0 47.64 00 E) and Urmia (37 ◦ 32 0 55 00 N and 45 ◦ 04 0 03 00 E). The characteristics of these
characteristics of these plants are summarized in Appendix A. One hundred and fifty
plants are summarized in Appendix A. One hundred and fifty
leaves for each medicinal plant (a total of 750 samples) were used for the investigation. leaves for each medicinal
plant
After (a
thetotal of 750 samples)
collection, the leaveswerewere
usedimmediately
for the investigation.
wrappedAfter the collection,
in zipped plastic the
bags leaves
and
were immediately wrapped in zipped plastic bags and
transferred to the laboratory for the following investigation operations. transferred to the laboratory for the
following investigation operations.
The leaves’ images were captured in an imaging box. The imaging box consists of a
The leaves’ images
camera, lighting box, andwere captured
computer andinisan imagingwith
equipped box. aThe ring imaging
of LEDbox consists
lamps of a
that emit
camera, lighting box, and computer and is equipped with a ring
low-intensity infrared light (450 lm) with adjustable light intensity on 3 levels (low, me- of LED lamps that emit
low-intensity infrared light (450 lm) with adjustable light intensity on 3 levels (low, medium,
dium, and high). Refer to Azadnia and Kheiralipour [13] for more details on the imaging
and high). Refer to Azadnia and Kheiralipour [13] for more details on the imaging box.
box. The distance for imaging the leaf samples was set at 250 mm. The images were cap-
The distance for imaging the leaf samples was set at 250 mm. The images were captured
tured with a smartphone (Galaxy A8, SAMSUNG Corporation, South Korea, 16 MP cam-
with a smartphone (Galaxy A8, SAMSUNG Corporation, Suwon, Korea, 16 MP camera).
era). The mobile camera settings were adjusted to achieve high-quality images (f/1.2, 9.8
The mobile camera settings were adjusted to achieve high-quality images (f/1.2, 9.8 mm).
mm). In addition, for each leaf sample, an RGB image with 3456 × 4608 pixels size was
In addition, for each leaf sample, an RGB image with 3456 × 4608 pixels size was taken
taken and stored as a jpeg file on a personal computer.
and stored as a jpeg file on a personal computer.
2.2. Image
2.2. Image Pre-Processing
Pre-Processing
Plant leaf images
Plant leaf images were
werepre-processed
pre-processedtotoremove
removebackgrounds
backgrounds using
using a code
a code written
written in
in the Python programming software environment (3.10v). The images
the Python programming software environment (3.10v). The images were automatically were automati-
cally uploaded
uploaded by thebyprogram
the program and processed.
and processed. OpenOpen CV2 library,
CV2 library, Python Python
ImageImage Library
Library (PIL),
(PIL), and Numpy library were exploited to build the code to process
and Numpy library were exploited to build the code to process the images. The optimal the images. The
T
optimal T value (threshold value for removing the background) was selected
value (threshold value for removing the background) was selected and applied to separate and applied
to separate
the the leaves
leaves from from the backgrounds.
the backgrounds. This Otsu’s
This study used study used Otsu’s
threshold tothreshold to identify
identify the optimal
the optimal
threshold threshold
value [34]. value [34].
The pre-processing operation for background removal was performed via the the follow-
follow-
steps are
ing steps, while the results of the steps are shown
shown in in Figure
Figure 1:
1:
Step 1: The acquired images were re-sized.
Step 2: The images were optimized by Otsu’s threshold method. method.
Step 3: Suppression of the empty pixels inside the leaf images via dilation operation
through the morphology method.
Step 4: Reversing the binary mask made in the previous step.
Step 5: Replacement of the reversed mask by the pixels related to the main images of
the plants.

Figure 1. Image pre-processing steps to remove the background from an original image (see text for
descriptions).
1–5 descriptions).

The speed
speed and
andaccuracy
accuracyofofthetheprocessing
processing ofof
thethe images
images through
through thethe
DL DL network
network are
are affected
affected mainly
mainly by pixel
by pixel image image
sizessizes
[35]. [35].
To beTo be effective,
effective, automatic
automatic visionvision systems
systems must
must
be fastbeand
fastaccurate.
and accurate. For
For this this reason,
reason, the images
the images were re-sized
were re-sized in threeindefinition
three definition
levels,
levels, 256 × 256, 128 × 128, and 64 × 64 pixels, to enhance the speed and investigate the
classification accuracy rate at the different image definitions. Generally, it is not necessary
to pre-process images to train deep learning algorithms. We explored the possible reduction
of the computational time and increase of the recognition accuracy by pre-processing the
images.

2.3. Data Augmentation (DA)

Data Augmentation (DA) is a technique to effectively increase the amount of data
for network training, addressed at enhancing the accuracy of models and significantly
 network training, addressed at enhancing the accuracy of models and significantly pre-
venting its over-fitting [36]. For the investigation, in addition to the original images, five
DA multiform transformation methods were adopted, as reported in Figure 2: four angles
rotations, 45°, 90°, 135°, and 180°, and a color manipulation. From the total number of data
(13,500 images), 80% (10,800 images) and 20% (2700 images) were randomly selected for
Agronomy 2022, 12, 2723 4 of 16
training and testing the proposed network, respectively (Table 1).

Table 1. Data splitting for training and testing in the proposed CNN.
preventing its over-fitting [36]. For the investigation, in addition to the original images, five
DA Set Data Splitting
multiform (%) Original
transformation methods Images Augmented
were adopted, Data
as reported Total
in Figure Images
2: four angles
Train ◦ ◦80 ◦ ◦ 600 10,200 10,800
rotations, 45 , 90 , 135 , and 180 , and a color manipulation. From the total number of data
Test images), 80%
(13,500 20 (10,800 images)150and 20% (2700 images)2550 were randomly2700 selected for
Total and testing
training 100 the proposed network,
750 respectively12,750
(Table 1). 13,500

Figure
Figure 2. The DA
2. The DA method
method using
using image
image rotation
rotation and
and color
color manipulation.
manipulation.

2.4. Architecture
Table of thefor
1. Data splitting Proposed
trainingCNN Modelin the proposed CNN.
and testing
CNN is an
Set
architecture of DLOriginal
Data Splitting (%)
algorithms containing
Images
a collection of non-linear
Augmented Data
trans-
Total Images
formation functions. These networks are an advanced version of artificial neural networks
Train 80 600 10,200 10,800
(ANNs) that can perform image processing operations such as object identification, seg-
Test 20 150 2550 2700
mentation,
Total and classification.
100 CNN models
750 are made up 12,750
of several layers producing
13,500 an
output from an input. Convolutional, Pooling, and Fully Connected layers are among the
2.4. Architecture of the Proposed CNN Model
CNN is an architecture of DL algorithms containing a collection of non-linear transfor-
mation functions. These networks are an advanced version of artificial neural networks
(ANNs) that can perform image processing operations such as object identification, segmen-
tation, and classification. CNN models are made up of several layers producing an output
from an input. Convolutional, Pooling, and Fully Connected layers are among the most
common algorithm of a CNN network implemented to process RGB images. Convolutional
layers use the local correlation of information in the images to extract the features: in other
words, they are applied to the input data to extract low-level features such as edges, corners
and blobs, and generate a feature map by using a set of filters, i.e., the so-called kernels [37].
Pooling layers select features from the top-layer feature map using sampling and holding
the model steady when performing operations, such as rotation and scaling. These two
layers are employed to reduce the network parameters and training time and prevent
overfitting [38]. The two most common types of pooling layers are maximum and average
pooling layers. The Fully Connected layers find an application after the convolutional and
pooling layers have been implemented. These layers include neurons, biases, and weights.
 as edges, corners and blobs, and generate a feature map by using a set of filters, i.e., the
so-called kernels [37]. Pooling layers select features from the top-layer feature map using
sampling and holding the model steady when performing operations, such as rotation
and scaling. These two layers are employed to reduce the network parameters and train-
Agronomy 2022, 12, 2723 ing time and prevent overfitting [38]. The two most common types of pooling layers 5 ofare
16
maximum and average pooling layers. The Fully Connected layers find an application
after the convolutional and pooling layers have been implemented. These layers include
neurons,
In biases,
the Fully and weights.
Connected In theneuron
layers, each Fully Connected
is connectedlayers,
to the each
upperneuron
neuronistoconnected
convert theto
the upper neuron to convert the integrated multidimensional features
integrated multidimensional features into one-dimensional features for classification and into one-dimen-
sional featuresoperations
identification for classification
[39]. and identification operations [39].
Figure 33 summarizes
Figure summarizesthe thearchitecture
architectureofofthe themodel
modeladopted
adopted forfor
image
image recognition.
recognition. It
consists
It consistsof of
5 convolutional
5 convolutional blocks
blocksandanda classifier
a classifierblock. In the
block. 5 convolutional
In the 5 convolutionalblocks, the
blocks,
output
the of each
output block
of each is the
block is input of the
the input of next one.one.
the next In each convolutional
In each convolutional block, 2 convolu-
block, 2 con-
tional layers are used to extract the important features, including shape,
volutional layers are used to extract the important features, including shape, color, and color, and texture.
Stride and
texture. Padding
Stride and of convolutional
Padding layers were
of convolutional 3 × 3were
layers kernels,
3 ×equivalent
3 kernels, to 1 pixel. An
equivalent to
1activation
pixel. Anfunction was
activation considered
function was for each ReLU
considered for layer. Each convolutional
each ReLU layer is fol-
layer. Each convolutional
lowedisby
layer a batchbynormalization
followed layer, which
a batch normalization layer,causes
which the CNNs
causes to become
the CNNs deeper,
to become thus
deeper,
reducing
thus the number
reducing the number of training
of training iterations
iterationsrequired.
required.After
Afterbatch
batchnormalization
normalization layers,
layers,
max-pooling layers of 2 × × 22 with
with Stride
Stride 22 were
were utilized
utilized to reduce the dimensions of the
feature map. Finally,
Finally, a dropout
dropout layer
layer with
with a value of 0.1 was used in each block to prevent
overfitting.

Figure 3. The architecture of the CNN model for image recognition.

recognition.

The output of the convolutional blocksblocks is the input of the following

following classifier
classifier block.
block.
The architecture
architectureof
ofthe
theclassifier
classifierisisdescribed
describedininFigure 4. 4.
Figure It It
contains 4 layers:
contains Global
4 layers: Average
Global Aver-
Pooling (GAP),
age Pooling dense
(GAP), (ReLU),
dense dropout,
(ReLU), and and
dropout, softmax layers.
softmax layers.
In this study, we adopted the GAP layer instead of a Fully Connected layer as reported
in similar experiments [40]. The main advantage of the GAP layer is to prevent the over-
fitting phenomenon by reducing the number of parameters and complexity of network
computing. The main difference between GAP and fully connected layers is that the output
of the flatten layer must be given to the fully connected layer to obtain the required features
(Figure 5). However, the output of the GAP layer has appropriate dimensions, so the fully
connected layer is not required. As a result, when fully connected layers are used, many
parameters are added to the model, resulting in a more complex and a higher probability
of over-fitting. Conversely, in the GAP layer there is no need for the parameters to be
optimized; therefore, the model reduces the number of parameters and has a low degree of
complexity.
Feature extraction of the GAP layer from the images is summarized in Figure 6. The
GAP layer reduces the spatial dimension of a tensor (N × M × D) to a tensor with the
dimension of 1 × 1 × D.
 Agronomy 2022, 12,
Agronomy 2022, 12, 2723
x FOR PEER REVIEW 6 of 16
15

Figure 4. A block diagram of the proposed CNN model.

In this study, we adopted the GAP layer instead of a Fully Connected layer as re-
ported in similar experiments [40]. The main advantage of the GAP layer is to prevent the
over-fitting phenomenon by reducing the number of parameters and complexity of net-
work computing. The main difference between GAP and fully connected layers is that the
output of the flatten layer must be given to the fully connected layer to obtain the required
features (Figure 5). However, the output of the GAP layer has appropriate dimensions, so
the fully connected layer is not required. As a result, when fully connected layers are used,
many parameters are added to the model, resulting in a more complex and a higher prob-
ability of over-fitting. Conversely, in the GAP layer there is no need for the parameters to
be optimized; therefore, the model reduces the number of parameters and has a low de-
gree of complexity.
Figure 4. A block diagram of the proposed CNN model.

In this study, we adopted the GAP layer instead of a Fully Connected layer as re-
ported in similar experiments [40]. The main advantage of the GAP layer is to prevent the
over-fitting phenomenon by reducing the number of parameters and complexity of net-
work computing. The main difference between GAP and fully connected layers is that the
output of the flatten layer must be given to the fully connected layer to obtain the required
features (Figure 5). However, the output of the GAP layer has appropriate dimensions, so
Agronomy 2022, 12, x FOR PEER REVIEW 7 of 15
the fully connected layer is not required. As a result, when fully connected layers are used,
many parameters are added to the model, resulting in a more complex and a higher prob-
ability 5.ofDifferences
Figure over-fitting. Conversely,
in the in of
performance the GAP
GAP vs.layer there is nolayers
fully connected needinfor the parameters
imaging to
classification.
Figure 5. Differences
be optimized; in the the
therefore, performance of GAP vs.
model reduces thefully connected
number layers in imaging
of parameters classifica-
and has a low de-
tion.
gree of complexity.
Feature extraction of the GAP layer from the images is summarized in Figure 6. The
GAP layer reduces the spatial dimension of a tensor (N × M × D) to a tensor with the
dimension of 1 × 1 × D.

Figure 6. The ability of the GAP layer to identify medicinal plants.

Figure
Figure 5. Differences
6. The in the
ability of the GAPperformance of GAP
layer to identify vs. fully plants.
medicinal connected layers in imaging classifica-
tion. The ReLU function in the classifier block performs mathematical operations
(Equation
The ReLU
Feature (1)) on the input
function
extraction in thedata,
ofthe GAP while
classifier the performs
block
layer from softmax function calculates
mathematical
the images inthe
operations
is summarized numerical
Figure(Equa-
6. The
value
tion of the normalized probability for each neuron based on (Equation (2))
GAP layer reduces the spatial dimension of a tensor (N × M × D) to a tensor withofthe
(1)) on the input data, while the softmax function calculates the numerical[41].
value
thedimension
normalized of probability
1 × 1 × D. for each neuron  based on (Equation (2)) [41].
x if x > 0
f ( x ) = 𝑥 𝑖𝑓 𝑥 > 0 (1)
𝑓(𝑥) = { 0 otherwise (1)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
exp(𝑎𝑖 )
𝑝𝑖 = 𝑛 (2)
∑𝑗=1 exp(𝑎𝑗 )

where 𝑎𝑖 is the input of the softmax function for node i (Class i).
 Figure 6. The ability of the GAP layer to identify medicinal plants.

The ReLU function in the classifier block performs mathematical operations (Equa-
tion (1)) on the input data, while the softmax function calculates the numerical value of
the normalized probability for each neuron based on (Equation (2)) [41].
Agronomy 2022, 12, 2723 7 of 16
𝑥 𝑖𝑓 𝑥 > 0
𝑓(𝑥) = { (1)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
exp
exp ( a(𝑎
i )𝑖 )
pi 𝑝𝑖
== n 𝑛 (2)
∑
∑ j=𝑗=1
1 exp a(𝑎
exp j 𝑗)

where 𝑎a𝑖i is the input of the softmax function for node i (Class i).
where
Figure 7 summarizes the classification model adopted to classify the leaf images. images. It
includes a GAP layer,
layer, 33 dense
denselayers
layerswith
with64,64,128,
128,and
and256
256neurons,
neurons,a adropout
dropoutlayer,
layer,and
anda
softmax
a softmax classifier. Finally,
classifier. the the
Finally, classifier block
classifier withwith
block threethree
different densedense
different layers layers
were tested
were
to select
tested tothe model
select with the
the model bestthe
with performance.
best performance.

Figure 7. A summary of the proposed model for the identification of medicinal

medicinal plant
plant species.
species.

2.5. Performance Metrics

2.5. Performance Metrics
A
A confusion
confusionmatrix
matrixwas
wasutilized toto
utilized evaluate
evaluatethethe
predictive performance
predictive performanceof the
oftest
thedata
test
after the classification. The confusion matrix is a table layout that allows for visualizing
data after the classification. The confusion matrix is a table layout that allows for visual- the
performance of a supervised algorithm. Thus, the performance metrics, including
izing the performance of a supervised algorithm. Thus, the performance metrics, includ- accuracy,
precision, sensitivity, specificity, and Area-Under-the-Curve (AUC) were measured based
ing accuracy, precision, sensitivity, specificity, and Area-Under-the-Curve (AUC) were
on the confusion matrix elements.
measured based on the confusion matrix elements.
3. Results
3. Results
The most common way to visually assess the model’s performance is to extract the
Theusing
features mostacommon way to filter
visual response visually
[42].assess the model’s
The proposed CNNperformance is to
model extracts theextract
low-levelthe
Agronomy 2022, 12, x FOR PEER REVIEWfeatures using a visual response filter [42]. The proposed CNN model extracts the low-
8 of 15
features such as edges, blobs and corners from the images by the initial layers, while
level features
high-level such as
features edges,
such blobs and
as textures andcorners
objectsfrom the images
are identified byby the
the initial
deep layers,
layers. while
Figure 8
high-level features such as textures and objects are identified by the deep
summarizes how the proposed CNN model extracts low-level features from the images. layers. Figure 8
summarizes
These
These features
features how the
contain
contain proposed CNN model
efficientinformation
efficient informationwhichextracts
whichthe low-level
theproposed features
proposedmodel from
model extracts the images.
extracts in the initial
in-
itiallayers.
layers.

Figure 8. The
Figure firstfirst
8. The activation layer
activation of each
layer channel
of each related
channel to the
related testtest
to the images.
images.

TheThe result
result of of
thethe extractionofofthe
extraction thefeatures
featuresand
andthe
thefollowing
following classification
classification of
ofthe
theimages
im-
of of
ages thethe
medicinal
medicinal plants
plantsthrough
throughthe theCNN
CNNnetwork
networkadopted
adoptedfor
forthe
the study
study is presented in
is presented
the confusion
in the confusion matrix
matrix reported
reported inin Figure
Figure 9.
9.
 The result of the extraction of the features and the following class
ages of the medicinal plants through the CNN network adopted for the
Agronomy 2022, 12, 2723 8 of 16
in the confusion matrix reported in Figure 9.

Figure 9. Predicted distributions of the 5 classes of medicinal plants images based on their sizes of
(a) 64 × 64, (b) 128 × 128, and (c) 256 × 256 pixels.

The proposed CNN model has been able to correctly classify the images of the medici-
nal plants with overall accuracy rates of 99.66, 99.32, and 99.45% when the input images are
 Figure 9. Predicted distributions of the 5 classes of medicinal plants images based on their
Agronomy 2022, 12, 2723 9 of 16
(a) 64 × 64, (b) 128 × 128, and (c) 256 × 256 pixels.

The proposed CNN model has been able to correctly classify the images of t
64 × 64, 128 × 128, and 256 × 256 pixels, respectively. Finally, according to the results, the
dicinal confusion
plants with matrixoverall
of 64×64accuracy
pixels achievedratestheof 99.66,
best 99.32,with
performance andan99.45% when the inp
overall accuracy
ages areof 99.66%.
64 × 64,Moreover,
128 × 128, in theand
confusion
256 ×matrix of 64 × 64
256 pixels, pixels, the Peppermint
respectively. Finally,class
according
had a mistaken image, which is confused with plants in the Tulsi class. The Bael and
results,Stevia
the confusion matrix of 64×64 pixels achieved the best performance with an
classes had three misclassified images. The Tulsi class had two misclassified images,
accuracywhichof 99.66%.
was confused Moreover, in the confusion
with the Peppermint and Steviamatrix
classes. of
This64could
× 64bepixels,
attributedthe Pepp
class had a mistaken
to the similarities image,
in shape which
features is confused
among them. In with plants
particular, theinLemon
the Tulsi
Balm class.
leaves, The B
in all three pixel dimensions, were classified correctly by the CNN model without any
Stevia classes had three misclassified images. The Tulsi class had two misclassifi
confusion in all images (100% accuracy). However, the high similarities among the leaves
ages, which was confused
of medicinal plants caused with the Peppermint
the proposed and Stevia
model to incorrectly classes.
classify This could
several images of the be attr
to the similarities
other 4 speciesin shapeinfeatures
involved the study. among
Therefore,them.
althoughInthe
particular,
medicinal planttheimages
Lemon wereBalm lea
similar, the model nearly predicted the correct classifications.
all three pixel dimensions, were classified correctly by the CNN model without an
Figure 10 indicates the CNN model’s prediction accuracy and loss function for the
fusion in
trainalland
images
test leaf(100%
images accuracy). However,
with three different the high
image sizes. similarities
The accuracy among
and loss values the le
medicinal plants for
are obtained caused the proposed
each epoch. Each epoch goes model to incorrectly
through a cycle to update classify several
its weight during images
total data training. The loss values determine how much the proposed model reacts after
other 4 species involved in the study. Therefore, although the medicinal plant image
optimizing each iteration. In other words, the loss values train the sum of errors created
similar,forthe
eachmodel
image to nearly
find the predicted
model’s bestthe correct
weights. Theclassifications.
downward trend in the training and
Figure 10 indicates the CNN model’s prediction
testing diagrams shows that the proposed model has excellentaccuracy
performanceand loss function
in classifying
the medicinal plants’ images. Convergence of the proposed CNN model was obtained after
train and test leaf images with three different image sizes. The accuracy and loss
100 epochs; in other similar models, with hundreds of classes and millions of images, it
are obtained
could befor each
close to 1000epoch.
epochsEach epoch
[43]. The goes
incorrect through
adjustment a cycle tosuch
of parameters, update its weight
as learning
total data
rate,training.
in the network,Thecauses
loss values
the model determine how
to be over-fitted andmuch
fail to the
reachproposed
convergence. model reac
Table 2 summarizes the accuracy, precision, sensitivity, specificity, and AUC of the
optimizing each iteration. In other words, the loss values train the sum of errors c
three sizes of images extracted by the proposed CNN confusion matrix. All of them are
for eachat image
100% fortothe findLemonthe Balm
model’simages.bestTheweights. The downward
highest average per class value trend in the traini
of accuracy
testing (99.8%)
diagrams showsfor
was achieved that
thethe
image proposed
size of 64 ×model has
64 pixels. excellent
Moreover, performance
the average values of in clas
accuracy for 128 × 128 and 256 × 256 pixel images were 99.7%. The results also indicate
the medicinal plants’ images. Convergence of the proposed CNN model was ob
that the highest and lowest precision metric values were obtained for the image size of
after 100
64 epochs;
× 64 and 128 in other similar
× 128 pixels withmodels, with respectively.
99.6 and 99.2%, hundreds According
of classestoand Tablemillions
2, the of i
it couldhighest
be close to 1000
average epochs
per class value[43].
for AUCThewasincorrect
obtainedadjustment
for images of 64 of×parameters,
64 pixels with such a
ing rate, in the network, causes the model to be over-fitted and failofto
an accuracy of 99.7%. The values of this parameter for images with the sizes × 128 conver
128reach
and 256 × 256 pixels were slightly lower at 99.5 and 99.6%, respectively. However, the
specificity values were 99.8% for the three sizes of images.

(a)
Figure 10. Cont.
nomy 2022, 12, x FOR PEER REVIEW 10 of
Agronomy 2022, 12, 2723 10 of 16

(b)

(c)
FigureFigure
10. Accuracy
10. Accuracyandand
loss performance
loss performance in inthe
the training
training andand testing
testing phasesphases for image
for the leaf the leaf image siz
sizes
of (a) 64 × 64,
of (a) (b)64,128
64 × (b) × 128,
128 andand
× 128, (c)(c)
256256××256
256pixels obtained
pixels obtained through
through the proposed
the proposed CNN model.
CNN model.

Table 2 summarizes the accuracy, precision, sensitivity, specificity, and AUC of th

three sizes of images extracted by the proposed CNN confusion matrix. All of them are
100% for the Lemon Balm images. The highest average per class value of accuracy (99.8%
was achieved for the image size of 64 × 64 pixels. Moreover, the average values of accura
for 128 × 128 and 256 × 256 pixel images were 99.7%. The results also indicate that th
highest and lowest precision metric values were obtained for the image size of 64 × 64 an
128 × 128 pixels with 99.6 and 99.2%, respectively. According to Table 2, the highest ave
age per class value for AUC was obtained for images of 64 × 64 pixels with an accuracy
99.7%. The values of this parameter for images with the sizes of 128 × 128 and 256 × 2
pixels were slightly lower at 99.5 and 99.6%, respectively. However, the specificity valu
were 99.8% for the three sizes of images.
Agronomy 2022, 12, 2723 11 of 16

Table 2. The performance results of medicinal plant classification obtained from the proposed CNN model.

Performance Classification Indicators

Plant Samples Image Size of 64 × 64 Image Size of 128 × 128 Image Size of 256 × 256
Accuracy Precision Sensitivity Specificity AUC Accuracy Precision Sensitivity Specificity AUC Accuracy Precision Sensitivity Specificity AUC
Lemon Balm 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
Peppermint 0.998 0.997 0.996 0.999 0.997 0.997 0.993 0.991 0.998 0.994 0.997 0.991 0.995 0.998 0.996
Bael 0.998 0.994 0.998 0.998 0.998 0.995 0.983 0.992 0.995 0.993 0.996 0.996 0.990 0.999 0.994
Stevia 0.998 0.995 0.993 0.999 0.996 0.996 0.991 0.991 0.998 0.994 0.996 0.989 0.991 0.997 0.994
Tulsi 0.997 0.994 0.996 0.998 0.997 0.997 0.990 0.996 0.999 0.997 0.997 0.993 0.996 0.998 0.997
Average per
0.998 0.996 0.996 0.998 0.997 0.997 0.992 0.994 0.998 0.995 0.997 0.994 0.994 0.998 0.996
class
Agronomy 2022, 12, 2723 12 of 16

4. Discussion
According to the results, the proposed CNN algorithm has been proven to be effective
in identifying medicinal plants with similar leaves compared to models developed by
previous studies for a similar purpose.
Amuthalingeswaran et al. [44] implemented a DL model to identify medicinal plants.
Their model was trained on 800 images of 4 types of medicinal plants and could identify
medicinal plants in the field with 85% accuracy. Anubha Pearline et al. [45] identified
different plants using DL methods and conventional ML algorithms. The plant classification
was performed by conventional learning methods extracting texture, shape and color
features using LBP and Haralick algorithms, the HU moments, and various color channels,
respectively. The extracted features were then classified by Linear Discriminant Analysis
(LDA), Logistic Regression (LR), K-Nearest Neighbor (KNN), Classification and Regression
Tree (CART), and Random Forest (RF). The best identification performance, 82.38%, was
obtained with the RF algorithm. Furthermore, the VGG16 network identified plants with
97.14% accuracy. Zhu et al. [46] proposed a two-way attention model based on the DL
network to identify plants from the Flower 102 dataset. The recognition accuracy of their
model was 97.2%. Munner and Fati [47] recognized Malaysuan herbs using an automated
classification system. They extracted the shape and texture features of plant images and
classified them using a DL model. The performance metrics of accuracy, precision and
sensitivity achieved were 98, 93 and 85%, respectively. Moreover, Reddy et al. [48] proposed
an optimized CNN model consisting of four convolutional layers, followed by two fully
connected layers and a softmax layer. Their suggested CNN model was focused on the
color images of leaves. The accuracy, precision and sensitivity metrics obtained were 97.6,
93.4 and 95.2%, respectively. Moreover, Zhang et al. [49] proposed a 7-layer CNN approach
to classify 32 plant species. They employed the DA technique to increase the dataset images,
achieving a 94.7% accuracy rate in plant identification.
The results of the study highlight that, unlike conventional ML methods, the proposed
DL model can automatically extract important features from medicinal plant leaves and
classify them with an accuracy up to 99%. Hence, there is no need to extract the features
manually. Furthermore, experimental results indicated that our approach outperformed
previous studies for identifying medicinal plants in terms of precision, sensitivity, specificity,
and AUC by 99.6, 99.6, 99.8, and 99.7%, respectively.
Nonetheless, using primary CNN models has been associated with challenges, such
as high computational cost, complexity, and a long-time running process. To overcome this
problem, we adopted two strategies: (1) increasing the number of images and reducing
their sizes during image processing operations and (2) replacing the fully connected layer
with the GAP layer, which significantly decreased the number of parameters and model
complexity. Such solutions, finally, have been proven to increase the model’s accuracy
and computational speed and prevent the occurrence of the over-fitting phenomenon. In
addition, it is noteworthy that the prediction time for the collected medicinal plant leaves
database using the proposed model was 40.81 s, which is lower than the time, 44.1 s, for
the same operation reported in the study conducted by Roopashree and Anitha [50].
According to the results we obtained from the study, the method we adopted can
be used to design an automated vision-based recognition system to identify various herb
plants with high accuracy and speed. Such an application can improve the public’s interest
and use of medical herbs, supporting the demand for healthier and safer food. Furthermore,
the model developed by the study will be tested in future studies, with possible adjustments,
to identify less common medicinal plants.

5. Conclusions
Identifying medicinal plants separately from other non-edible plants is essential in
botany and food industries. The traditional methods of identifying medicinal plants are
time-consuming, complex, and require experienced and trained people. The automatic
real-time vision-based system exploited to identify broadly used medical herbs with similar
 Agronomy 2022, 12, 2723 13 of 16

leaves has given positive results. The proposed method includes an improved CNN
network consisting of convolutional and classifier blocks. The classifier had a Global
Average Pooling (GAP) layer, dense layer, dropout layer, and softmax layer. Compared
with previous studies’ results, this solution reduces the number of parameters and increases
the speed and accuracy of the model. The proposed CNN model identified the medicinal
plant images in three levels of image definition, 64 × 64, 128 × 128, and 256 × 256 pixels,
with overall accuracy rates of 99.66, 99.32, and 99.45%, respectively. Therefore, combining
image processing and the proposed CNN algorithm is an efficient alternative to traditional
methods. Future works will be carried out to improve the model’s performance in the
classification of further species of medicinal plants to confirm the effectiveness to a larger
extent of the solution we developed. The model will be adopted to develop an intelligent
mobile-based application for the real-time identification of medicinal plants.
The current study aimed to improve the automatic identification of medicinal plants
due to their growing popularity and increased requests for artisanal and industrial uses
and applications. Therefore, the proposed DL algorithm and image processing technique
can have a special place in plant science and even industrial markets for identifying and
classifying varied medicinal plants separately from other non-edible plants.

Author Contributions: Conceptualization, R.A.; methodology, R.A. and H.M.; software, R.A. and
M.A.C.; validation, R.A. and H.M.; formal analysis, R.A. and M.A.C.; investigation, R.A. and E.C.;
resources, M.A.C., A.D. and M.M.A.-A.; data curation, A.D. and M.M.A.-A.; Writing—original draft
preparation, R.A. and H.M.; Input to different sections of the manuscript and contribution to data
interpretation E.C.; Writing, Review, Editing and Finalization of the manuscript E.C.; visualization,
R.A. and M.A.C.; supervision, R.A. and E.C.; project administration, R.A.; funding acquisition, E.C.
All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Agronomy 2022, 12, x FOR PEER REVIEW 13 ofthat
Conflicts of Interest: We declare that we do not have any commercial or associative interest 15
Agronomy 2022, 12, x FOR PEER REVIEW
represent conflict of interest in connection with the work submitted. 13 of 15

AppendixAA
Appendix
Appendix
The A
Themedicinal
medicinalplants
plantsstudied
studiedand
andtheir
theirpertinent
pertinentdetails.
details.
Scientific Name CommonThe medicinal plants studiedMedicinal
Name and their pertinent
Propertiesdetails. Leaf Image
Scientific
Scientific Name
Name Common
Common Name Name Medicinal
Medicinal Properties
Properties
This type of medicinal plant belongs to the Lamiaceae fam- Leaf
Leaf Image
Image
This type
ily. This of type
This
herb medicinal
plant plant plant
of medicinal
contains belongs
belongs
a variety toofthe
to Lamiaceae
the Lamiaceae
secondary fam-
metabo-
ily. This family.
herb This herb
plant plant a
contains contains
variety a of
variety of secondary
secondary metabo-
Melissa officinalis L. Lemon Balm lites that are very beneficial to human health [51,52]. In tradi-
metabolites that are very beneficial to human health
Melissa officinalis
Melissa L.L. Lemon
officinalis BalmBalm
Lemon lites thatmedicine,
tional are very beneficial toused
human
this herbmedicine,
is health
tothis
treat [51,52]. Inmental
headaches, tradi-
[51,52]. In traditional herb is used to treat
tional medicine,
disorders, this herb
rheumatism
headaches, is used
mental andto treat headaches,
hypersensitivity
disorders, rheumatism and mental
[53].
disorders, rheumatism and hypersensitivity
hypersensitivity [53]. [53].
These plants belong to the Asteraceae family. Stevia can be a
These These belong
goodplants plants
substitute forbelong
to thetoAsteraceae
sugar the Asteraceae
source family.
family.
for diabetics. ItStevia
Stevia canbe
can
is also a
used
Stevia rebaudiana Bertoni Stevia be a good substitute for sugar source for diabetics. It is
Stevia rebaudiana Bertoni Stevia good substitute
to reduce for sugar source for diabetics. It is also used
Stevia rebaudiana Bertoni Stevia alsothe
usedblood glucose
to reduce levels,
the blood blood levels,
glucose pressure and
blood to
to reduce the blood
treat glucose
mental levels,
problems blood pressure
[54,55].
pressure and to treat mental problems [54,55]. and to
treat mental problems [54,55].
The Mentha balsamea plant is a multipurpose herb plant
The
fromMentha balsamea
the Lamiaceae plant This
family. is a multipurpose
plant is used in herb plant
Iran as an
from the Lamiaceae
antiviral, invigorating, family. This plant
stimulant is used in Iran
and antifungal agentas[56].
an
Mentha balsamea Wild Peppermint
antiviral, invigorating,
The extracted oil from stimulant and antifungal
Mentha balsamea is alsoagent
used[56].
as a
Mentha balsamea Wild Peppermint
The extracted
medicine oil from
to treat cancer,Mentha
colds,balsamea
sore throats, is also used tooth-
nausea, as a
medicine to treat achescancer, colds, sore
and muscle throats,
soreness [57].nausea, tooth-
aches and muscle soreness
Aegle marmelos is a fruit with purely medicinal properties.[57].
Aegle marmelos
The leaves of thisis fruit
a fruit with
are purely
feverish and medicinal
expectorant properties.
and are
Aegle marmelos Bael The
usedleaves of this fruit
for bleeding, are feverish
diarrhea and expectorant
and intestinal disorders.and It isare
also
 tional medicine, this herb is used to treat headaches, mental
tionaldisorders,
medicine, rheumatism
this herb is used to treat headaches,
and hypersensitivity mental
[53].
disorders, rheumatism and hypersensitivity [53].
These plants belong to the Asteraceae family. Stevia can be a
These plants belong
good substitute to thesource
for sugar Asteraceae family. Stevia
for diabetics. canused
It is also be a
Stevia rebaudiana
Agronomy 2022, 12,Bertoni
2723 Stevia good substitute for sugar source for diabetics. It is also used 14 of 16
Stevia rebaudiana Bertoni Stevia to reduce the blood glucose levels, blood pressure and to
to reduce the treatbloodmental
glucose levels, blood
problems [54,55]. pressure and to
treat mental problems [54,55].
Scientific Name Common Name The Mentha balsamea plant is aProperties
Medicinal multipurpose herb plant Leaf Image
The Mentha
from the balsamea plant is a multipurpose
is used inherb plant
TheLamiaceae
Mentha balsameafamily.plantThisis plant
a multipurpose Iranplant
herb as an
from the
antiviral,
from
Lamiaceae
invigorating,
the Lamiaceae
family. This
stimulant
family. This
plant
andplantis used
antifungal in
is used in
Iran
agent as[56].
Iran as
an
Mentha balsamea Wild Peppermint antiviral, invigorating, stimulant and antifungal
an antiviral, invigorating, stimulant and antifungal agent [56].
Mentha balsamea
Mentha Wild
balsamea Wild Peppermint The extracted oil from Mentha balsamea is also used as a
Peppermint
The extracted
agent
medicine to treat oilThe
[56]. from Mentha
extracted
cancer, balsamea
colds,oilsorefromthroats,
Mentha is also usedtooth-
balsamea
nausea, as
is a
medicinealso usedcancer,
to treat as a medicine
colds, to treatthroats,
sore cancer, colds,
nausea, sore
tooth-
throats,aches
nausea,and muscle soreness
toothaches and muscle [57].
soreness [57].
Aegle marmelos aches and muscle
is a fruitiswith soreness
purely [57].
medicinal properties.
Aegle marmelos a fruit with purely medicinal
Aegle marmelos
The leaves of thisisfruit
properties. aThe
fruit with
leaves
are ofpurely
feverish this and medicinal
fruit are feverish
expectorant properties.
and
and are
Aegle marmelos The leaves
Bael Bael used of this fruit
expectorant
for bleeding, andare
diarrheaare feverish
used and expectorant
for bleeding,
and intestinal diarrhea and
disorders. and are
It is also
Aegle marmelos
Aegle marmelos Bael used intestinal disorders. It is also used to treat urinary
usedfor bleeding,
to treat urinary diarrhea and intestinal
problems, regulate heart
problems, regulate heart rate and treat stomach ache
disorders.
rate andIt istreat
also
used to treat urinary problems,
stomach ache regulate
[58,59]. heart rate and treat
[58,59].
This plant stomach
belongs to ache
the [58,59]. family. It grows
Lamiaceae
This plant belongs to the Lamiaceae family. It grows mostly
This plant
in warm belongs
mostly
regions to the
in warm
and Lamiaceae
regions
is used andas an is family.
used as an
antiseptic, It grows mostly
antiseptic,
antiemetic,
Ocimum sanctum
Ocimum L.L.
sanctum TulsiTulsi antiemetic,
in warm regions anti-flatulence,
and is treatment sedative,
used as anofantiseptic, treatment of
antiemetic,
Ocimum sanctum L. Tulsi anti-flatulence, sedative,
gastrointestinal disorders, rheumatismgastrointestinal
and skin disor-
anti-flatulence, sedative,
ders, rheumatismdisorders treatment disorders [60,61]. disor-
of
and skin[60,61]. gastrointestinal
ders, rheumatism and skin disorders [60,61].
References
References
References
1.
1. Naserifar,
Naserifar, R.;
R.;R.;Bahmani,
Bahmani, M.;
M.; Abdi,
Abdi, J.;
J.; Abbaszadeh,
Abbaszadeh, S.; Nourmohammadi,
S.; Nourmohammadi,
Nourmohammadi,G.A.; G.A.; Rafieian-Kopaei,
G.A.;Rafieian-Kopaei,
Rafieian-Kopaei, M.
M. A
A review
review of
of the
the most
most im-
im-
1. 1. Naserifar,
Naserifar,
portant
portant R.;
native
native
Bahmani,M.;
Bahmani,
medicinal
medicinal
M.;Abdi,
Abdi,J.;
plants
plants of
of
J.; Abbaszadeh,
Iran
Iran
Abbaszadeh, S.;
effective
effective on
on S.; Nourmohammadi,
leishmaniasis
leishmaniasis G.A.;
according
according to
toRafieian-Kopaei,
Iranian
Iranian
M.M.A review
ethnobotanical
ethnobotanical A review of the
of
references.
references.
most
the Int.
Int.
important
most J.
J. im-
Adv.
Adv.
nativenative
portant medicinal plantsplants
medicinal of Iranofeffective on leishmaniasis
Iran effective on leishmaniasisaccording to Iranian
according ethnobotanical
to Iranian references.
ethnobotanical Int. J. Adv.
references. Int.Biotechnol.
J. Adv.
Biotechnol.
Biotechnol. Res.
Res. 2017,
2017, 8,8, 1330–1336.
1330–1336.
Res. 2017,
Biotechnol. 8,
Res. 1330–1336.
2017, 8, 1330–1336.
2.
2. Altemimi,
Altemimi, A.; Lakhssassi,
A.; Lakhssassi, N.; Baharlouei,
N.; Baharlouei, A.;
A.; Watson,
Watson, D.G.;
D.G.; Lightfoot,
Lightfoot, D.A.
D.A. Phytochemicals:
Phytochemicals: Extraction,
Extraction, isolation,
isolation, and
and identifi-
identifi-
2. 2. Altemimi,
Altemimi,
cation
cation of
of A.;A.;
bioactive
bioactive
Lakhssassi,
Lakhssassi,
compounds
compounds N.;N.; Baharlouei,
Baharlouei,
from
from plant
plant A.;A.; Watson,D.G.;
Watson,
extracts.
extracts. Plants
Plants
D.G.;Lightfoot,
2017,
Lightfoot,D.A.
2017, 6,
6, 42.
42.
D.A.Phytochemicals:
Phytochemicals:Extraction,
Extraction,isolation,
https://doi.org/10.3390/plants6040042
https://doi.org/10.3390/plants6040042
isolation,and andidentifi-
identifica-
tion of
cation of bioactive
bioactive compounds
compounds from from plant
plant extracts. Plants 2017,
extracts. Plants 6, 42.
2017, 6, 42. https://doi.org/10.3390/plants6040042
[CrossRef] [PubMed]
3.
3. Naeem,
Naeem, S.;
S.; Ali,
Ali, A.; Chesneau,
A.;A.;
Chesneau, C.;
C.; Tahir,
Tahir, M.H.;
M.H.; Jamal,
Jamal, F.;
F.; Sherwani,
Sherwani, R.A.K.;
R.A.K.; Ul
UlUl Hassan,
Hassan, M.
M.M. The
The classification
classification of
of medicinal
medicinal plant
plant
3. 3. Naeem,
Naeem,
leaves
leaves S.; S.;
Ali,Ali,
based
based A.;
on
on
Chesneau,
Chesneau,
multispectral
multispectral C.;C.; Tahir,
Tahir,
and
and
M.H.;
M.H.;
texture
texture
Jamal,
Jamal,
featureF.;F.;
feature
Sherwani,
Sherwani,
using
using
R.A.K.;
R.A.K.;
machine
machine
Hassan,
Ullearning
Hassan,
learning M. TheThe
approach.
approach.
classification
classification
Agronomy
Agronomy of of medicinal
medicinal
2021,
2021, 11,
11,
plant
plant
263.
263.
leaves
leaves based on multispectral and texture feature using machine learning
based on multispectral and texture feature using machine learning approach. Agronomy 2021, 11, 263. approach. Agronomy 2021, 11, 263. [CrossRef]
https://doi.org/10.3390/agronomy11020263.
https://doi.org/10.3390/agronomy11020263.
Ozioma, E.O.J.; Chinwe, O.A.N. Herbal medicines in African traditional medicine. Herb. Med. 2019, 10, 191–214.
4. https://doi.org/10.3390/agronomy11020263.
4.
4. 5. Ozioma,
Ozioma, E.O.J.;
E.O.J.; Chinwe,
Chinwe, O.A.N.
O.A.N. Herbal
Herbal medicines
medicinesPlantsin
in African
African traditional medicine.
traditionalPeople
medicine. Herb.
Herb. Med.
Med. 2019,
2019, 10,
10, 191–214.
191–214.
4. Amenu,
Ozioma, E. Use
E.O.J.; and Management
Chinwe, O.A.N. Herbal of Medicinal
medicines by Indigenous
in African traditional medicine. of Ejaji Area
Herb. (Chelya
Med. 2019, Woreda)
10, 191–214. West Shoa, Ethiopia:
5.
5. Amenu,
Amenu, E.
E. Use
Use and
and Management
Management of
of Medicinal
Medicinal Plants
Plants by
by Indigenous
Indigenous People
People of
of Ejaji
Ejaji Area
Area (Chelya
(Chelya Woreda)
Woreda) West
West Shoa,
Shoa, Ethiopia:
Ethiopia:
5. An
Amenu, Ethnobotanical
E. Use and Approach.
Management Master’s
of Medicinal Thesis, Addis
Plants by Ababa University,
Indigenous People Addis
of EjajiAbaba,
Area Ethiopia,
(Chelya 2007.
Woreda) West Shoa, Ethiopia:
An
An Ethnobotanical
Ethnobotanical Approach.
Approach. Master’s
Master’s Thesis,
Thesis, Addis
Addis Ababa
Ababa University,
University, Addis
Addis Ababa,
Ababa, Ethiopia,
Ethiopia, 2007.
2007.
6. AnHu, R.; Lin, C.; Xu,
Ethnobotanical W.; Liu, Y.; Long, C. Ethnobotanical study on medicinal plants used by Mulam people in Guangxi, China. J.
6.
6. Hu,
Hu, R.;
R.; Lin,
Lin, C.;
Ethnobiol. C.; Xu, Approach.
Xu,
Ethnomed.
W.; Liu,
Liu, Y.;
W.; 2020, Y.;Master’s
16,
Long,
Long,
40. C. Thesis,
C. Addis Ababa
Ethnobotanical
Ethnobotanical
[CrossRef] [PubMed] studyUniversity,
study on
on medicinal
medicinal Addis
plants
plants Ababa,
usedEthiopia,
used by
by Mulam
Mulam 2007.
people
people in in Guangxi,
Guangxi, China.
China. J. J.
6. Hu, R.;
Ethnobiol.
Ethnobiol.Lin, C.;
Ethnomed.Xu,
Ethnomed. W.; Liu,
2020,
2020, Y.;
16,
16, Long,
40.
40. C. Ethnobotanical study on medicinal plants used by Mulam people in Guangxi, China. J.
7. Ethnobiol.
Crini, G.; Lichtfouse,
Ethnomed. 2020, E.;16,
Chanet,
40. G.; G.; Morin-Crini, N. Applications of hemp in textiles, paper industry, insulation and building
7.
7. Crini,
Crini, G.;
G.; Lichtfouse,
Lichtfouse, E.;
E.; Chanet,
Chanet, G.; Morin-Crini,
Morin-Crini, N.
N. Applications
Applications of
of hemp
hemp in in textiles,
textiles, paper
paper industry,
industry, insulation
insulation and
and building
building
7. materials,
Crini, G.; horticulture,
Lichtfouse, E.; animalG.;
Chanet, nutrition,
Morin-Crini,food andN. beverages, nutraceuticals,
Applications of hemp incosmetics cosmetics
textiles, paperand hygiene,insulation
industry, medicine,and agrochemistry,
building
materials,
materials, horticulture,
horticulture, animal
animal nutrition,
nutrition, food
food and
and beverages,
beverages, nutraceuticals,
nutraceuticals, cosmetics and
and hygiene,
hygiene, medicine,
medicine, agrochemistry,
agrochemistry,
energy production
materials, horticulture, and environment: A review. Environ. Chem. Lett. 2020, 18, 1451–1476.and[CrossRef]
energy
energy
8. energy
Nabavi,
production
production
S.F.; Di andanimal
and
Lorenzo,
nutrition,
environment:
environment:
A.; Izadi, A food
A
M.;
review.
review.andEnviron.
beverages,
Environ.
Sobarzo-Sánchez,
nutraceuticals,
Chem.
Chem.
E.;
Lett.
Lett. 2020,
Daglia, 2020,
M.;
cosmetics
18,
18,
Nabavi,
1451–1476.
1451–1476.
S.M.
hygiene, medicine, agrochemistry,
https://doi.org/10.1007/s10311-020-
https://doi.org/10.1007/s10311-020-
Antibacterial effects of cinnamon: From
01029-2.
01029-2. production and environment: A review. Environ. Chem. Lett. 2020, 18, 1451–1476. https://doi.org/10.1007/s10311-020-
farm
01029-2. to food, cosmetic and pharmaceutical industries. Nutrients 2015, 7, 7729–7748. [CrossRef]
8.
8. Nabavi,
Nabavi, S.F.;
S.F.; Di Lorenzo,
DiE.C.;
Lorenzo, A.;
A.; Izadi,
Izadi, M.;
M.; Sobarzo-Sánchez,
Sobarzo-Sánchez, E.;
E.; Daglia,
Daglia, M.;
M.; Nabavi,
Nabavi, S.M.
S.M.ofAntibacterial
Antibacterial effects
effects of
of cinnamon:
cinnamon: From
From
8. 9. Nabavi,
Chukwuma,S.F.; Di Soladoye,
Lorenzo, A.; M.O.;M.;
Izadi, Feyisola, R.T. Traditional
Sobarzo-Sánchez, E.; medicine
Daglia, and
M.; the future
Nabavi, S.M. medicinal Plants
Antibacterial in of
effects Nigeria.
cinnamon:J. Med. Plants
From
farm
farm to
to food,
food, cosmetic
cosmetic and
and pharmaceutical
pharmaceutical industries.
industries. Nutrients
Nutrients 2015,
2015, 7,
7, 7729–7748.
7729–7748. https://doi.org/10.3390/nu7095359.
https://doi.org/10.3390/nu7095359.
Stud.
farm 2015, 3,
to food, E.C.; 23–29.
cosmetic and pharmaceutical industries. Nutrients 2015, 7, 7729–7748. https://doi.org/10.3390/nu7095359.
9.
9. Chukwuma,
Chukwuma, E.C.; Soladoye,
Soladoye, M.O.;
M.O.; Feyisola,
Feyisola, R.T.
R.T. Traditional
Traditional medicine
medicine and the
the future
andneural future of
of medicinal
medicinal Plants
Plants in inofNigeria.
Nigeria. J.
J. Med.
Med.
Sladojevic, E.C.;
9. 10.Chukwuma,
Plants
Plants Stud.
Stud.
S.; Arsenovic,
2015,
2015, Soladoye,
3,
3, 23–29.
23–29.
M.; Anderla,
M.O.; A.; Culibrk,
Feyisola, R.T. D.; Stefanovic,
Traditional D. Deep
medicine and the networks
future of based recognition
medicinal Plants in plant diseases
Nigeria. J. Med. by
leaf Stud.
image2015, classification. Comput. Intell. Neurosci. 2016, 2016, 3289801. [CrossRef]
10. Plants
10. Sladojevic,
Sladojevic, S.;
S.;Misra, 3, 23–29.M.;
Arsenovic,
Arsenovic, M.; Anderla,
Anderla, A.;
A.; Culibrk,
Culibrk, D.;
D.; Stefanovic,
Stefanovic, D.
D. Deep
Deep neural
neural networks
networks based
based recognition
recognition of
of plant
plant diseases
diseases
10.11.Sladojevic,
by
Singh, V.;S.;
leaf
leaf image
byAgric. image
A.K. Detection
Arsenovic,
classification.
classification. M.; Anderla,
Comput.
Comput.
of plant
Intell.
Intell.
leaf diseases
A.; Culibrk,
Neurosci.
Neurosci. D.;
2016,
2016,
using image
Stefanovic,
2016,
2016, 3289801.
3289801.
segmentation
D. Deep neural networksand soft computing
based
https://doi.org/10.1155/2016/3289801.
https://doi.org/10.1155/2016/3289801. of plantInf.
techniques.
recognition Process.
diseases
2017, 4, 41–49. [CrossRef]
by leaf image classification. Comput. Intell. Neurosci. 2016, 2016, 3289801. https://doi.org/10.1155/2016/3289801.
12. Wäldchen, J.; Mäder, P. Plant species identification using computer vision techniques: A systematic literature review. Arch.
Comput. Methods Eng. 2018, 25, 507–543. [CrossRef] [PubMed]
13. Azadnia, R.; Kheiralipour, K. Recognition of leaves of different medicinal plant species using a robust image processing algorithm
and artificial neural networks classifier. J. Appl. Res. Med. Aromat. Plants 2021, 25, 100327. [CrossRef]
14. Wang, J.; Mo, W.; Wu, Y.; Xu, X.; Li, Y.; Ye, J.; Lai, X. Combined Channel Attention and Spatial Attention Module Network for
Chinese Herbal Slices Automated Recognition. Front. Neurosci. 2022, 16, 920820. [CrossRef] [PubMed]
15. Mukherjee, G.; Tudu, B.; Chatterjee, A. A convolutional neural network-driven computer vision system toward identification
of species and maturity stage of medicinal leaves: Case studies with Neem, Tulsi and Kalmegh leaves. Soft Comput. 2021, 25,
14119–14138. [CrossRef]
16. Bisen, D. Deep convolutional neural network based plant species recognition through features of leaf. Multimed. Tools Appl. 2021,
80, 6443–6456. [CrossRef]
17. Azadnia, R.; Jahanbakhshi, A.; Rashidi, S. Developing an automated monitoring system for fast and accurate prediction of soil
texture using an image-based deep learning network and machine vision system. Measurement 2022, 190, 110669. [CrossRef]
18. Apolo-Apolo, O.E.; Pérez-Ruiz, M.; Martínez-Guanter, J.; Valente, J. A cloud-based environment for generating yield estimation
maps from apple orchards using UAV imagery and a deep learning technique. Front. Plant Sci. 2020, 11, 1086. [CrossRef]
Agronomy 2022, 12, 2723 15 of 16

19. Ferentinos, K.P. Deep learning models for plant disease detection and diagnosis. Comput. Electron. Agric. 2018, 145, 311–318.
[CrossRef]
20. Ziyaee, P.; Farzand Ahmadi, V.; Bazyar, P.; Cavallo, E. Comparison of Different Image Processing Methods for Segregation of
Peanut (Arachis hypogaea L.) Seeds Infected by Aflatoxin-Producing Fungi. Agronomy 2021, 11, 873. [CrossRef]
21. Yamashita, R.; Nishio, M.; Do, R.K.G.; Togashi, K. Convolutional neural networks: An overview and application in radiology.
Insights Imaging 2018, 9, 611–629. [CrossRef]
22. 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]
23. Russel, N.S.; Selvaraj, A. Leaf species and disease classification using multiscale parallel deep CNN architecture. Neural Comput.
Appl. 2022, 34, 19217–19237. [CrossRef]
24. Bodhwani, V.; Acharjya, D.; Bodhwani, U. Deep Residual Networks for Plant Identification. Procedia Comput. Sci. 2019, 152,
186–194. [CrossRef]
25. Lee, S.H.; Chan, C.S.; Mayo, S.J.; Remagnino, P. How deep learning extracts and learns leaf features for plant classification. Pattern
Recognit. 2017, 71, 1–13. [CrossRef]
26. 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]
27. Bedi, P.; Gole, P. Plant disease detection using hybrid model based on convolutional autoencoder and convolutional neural
network. Artif. Intell. Agric. 2021, 5, 90–101. [CrossRef]
28. Mohanty, S.P.; Hughes, D.P.; Salathé, M. Using Deep Learning for Image-Based Plant Disease Detection. Front. Plant Sci. 2016, 7,
1419. [CrossRef]
29. Barbedo, J.G.A. Plant disease identification from individual lesions and spots using deep learning. Biosyst. Eng. 2019, 180, 96–107.
[CrossRef]
30. Nasiri, A.; Taheri-Garavand, A.; Fanourakis, D.; Zhang, Y.-D.; Nikoloudakis, N. Automated Grapevine Cultivar Identification via
Leaf Imaging and Deep Convolutional Neural Networks: A Proof-of-Concept Study Employing Primary Iranian Varieties. Plants
2021, 10, 1628. [CrossRef]
31. Paulson, A.; Ravishankar, S. AI Based Indigenous Medicinal Plant Identification. In Proceedings of the 2020 Advanced Computing
and Communication Technologies for High Performance Applications (ACCTHPA), Cochin, India, 2–4 July 2020; pp. 57–63.
[CrossRef]
32. Grinblat, G.L.; Uzal, L.C.; Larese, M.G.; Granitto, P.M. Deep learning for plant identification using vein morphological patterns.
Comput. Electron. Agric. 2016, 127, 418–424. [CrossRef]
33. Hu, J.; Chen, Z.; Yang, M.; Zhang, R.; Cui, Y. A Multiscale Fusion Convolutional Neural Network for Plant Leaf Recognition.
IEEE Signal Process. Lett. 2018, 25, 853–857. [CrossRef]
34. Vala, H.J.; Baxi, A. A review on Otsu image segmentation algorithm. Int. J. Adv. Res. Comput. Eng. Technol. IJARCET 2013, 2,
387–389.
35. Jahanbakhshi, A.; Momeny, M.; Mahmoudi, M.; Zhang, Y.-D. Classification of sour lemons based on apparent defects using
stochastic pooling mechanism in deep convolutional neural networks. Sci. Hortic. 2020, 263, 109133. [CrossRef]
36. Shorten, C.; Khoshgoftaar, T.M. A survey on Image Data Augmentation for Deep Learning. J. Big Data 2019, 6, 60. [CrossRef]
37. Ma, J.; Jiang, X.; Fan, A.; Jiang, J.; Yan, J. Image matching from handcrafted to deep features: A survey. Int. J. Comput. Vis. 2021,
129, 23–79. [CrossRef]
38. Farooq, M.; Sazonov, E. Feature Extraction Using Deep Learning for Food Type Recognition. Bioinform. Biomed. Eng. 2017, 10208,
464–472. [CrossRef]
39. Gu, J.; Wang, Z.; Kuen, J.; Ma, L.; Shahroudy, A.; Shuai, B.; Liu, T.; Wang, X.; Wang, G.; Cai, J.; et al. Recent advances in
convolutional neural networks. Pattern Recognit. 2018, 77, 354–377. [CrossRef]
40. Chen, Y.; Peng, G.; Xie, C.; Zhang, W.; Li, C.; Liu, S. ACDIN: Bridging the gap between artificial and real bearing damages for
bearing fault diagnosis. Neurocomputing 2018, 294, 61–71. [CrossRef]
41. Tang, Y. Deep learning using linear support vector machines. arXiv 2013, arXiv:1306.0239. [CrossRef]
42. Nasiri, A.; Taheri-Garavand, A.; Zhang, Y.-D. Image-based deep learning automated sorting of date fruit. Postharvest Biol. Technol.
2019, 153, 133–141. [CrossRef]
43. Azizi, A.; Gilandeh, Y.A.; Mesri-Gundoshmian, T.; Saleh-Bigdeli, A.A.; Moghaddam, H.A. Classification of soil aggregates: A
novel approach based on deep learning. Soil Tillage Res. 2020, 199, 104586. [CrossRef]
44. Amuthalingeswaran, C.; Sivakumar, M.; Renuga, P.; Alexpandi, S.; Elamathi, J.; Hari, S.S. Identification of Medicinal Plant’s and
Their Usage by Using Deep Learning. In Proceedings of the 2019 3rd International Conference on Trends in Electronics and
Informatics (ICOEI), Tirunelveli, India, 23–25 April 2019; pp. 886–890. [CrossRef]
45. Pearline, S.A.; Kumar, V.S.; Harini, S. A study on plant recognition using conventional image processing and deep learning
approaches. J. Intell. Fuzzy Syst. 2019, 36, 1997–2004. [CrossRef]
46. Zhu, Y.; Sun, W.; Cao, X.; Wang, C.; Wu, D.; Yang, Y.; Ye, N. TA-CNN: Two-way attention models in deep convolutional neural
network for plant recognition. Neurocomputing 2019, 365, 191–200. [CrossRef]
47. Muneer, A.; Fati, S.M. Efficient and automated herbs classification approach based on shape and texture features using deep
learning. IEEE Access 2020, 8, 196747–196764. [CrossRef]
Agronomy 2022, 12, 2723 16 of 16

48. Reddy, S.R.; Varma, G.P.; Davuluri, R.L. Optimized convolutional neural network model for plant species identification from leaf
images using computer vision. Int. J. Speech Technol. 2021, 1–28. [CrossRef]
49. Zhang, C.; Zhou, P.; Li, C.; Liu, L. A convolutional neural network for leaves recognition using data augmentation. In
Proceedings of the 2015 IEEE International Conference on Computer and Information Technology; Ubiquitous Computing and
Communications; Dependable, Autonomic and Secure Computing; Pervasive Intelligence and Computing, Liverpool, UK, 26–28
October 2015; pp. 2143–2150. [CrossRef]
50. Roopashree, S.; Anitha, J. DeepHerb: A Vision Based System for Medicinal Plants Using Xception Features. IEEE Access 2021, 9,
135927–135941. [CrossRef]
51. Mimica-Dukic, N.; Bozin, B.; Sokovic, M.; Simin, N. Antimicrobial and Antioxidant Activities of Melissa officinalis L. (Lamiaceae)
Essential Oil. J. Agric. Food Chem. 2004, 52, 2485–2489. [CrossRef]
52. Tagashira, M.; Ohtake, Y. A New Antioxidative 1,3-Benzodioxole from Melissa officinalis. Planta Med. 1998, 64, 555–558. [CrossRef]
53. Reiter, M.; Brandt, W. Relaxant effects on tracheal and ileal smooth muscles of the guinea pig. Arzneim. Forsch. 1985, 35, 408–414.
54. Rincon, F.; Mayer, S.A. Intracerebral Hemorrhage: Clinical Overview and Pathophysiologic Concepts. Transl. Stroke Res. 2012, 3,
10–24. [CrossRef]
55. Madan, S.; Ahmad, S.; Singh, G.N.; Kohli, K.; Kumar, Y.; Singh, R.; Garg, M. Stevia rebaudiana (Bert.) Bertoni—A review. Indian J.
Nat. Prod. Resour. 2010, 1, 267–286.
56. Herro, E.; Jacob, S.E. Mentha piperita (Peppermint). Dermatitis 2010, 21, 327–329. [CrossRef] [PubMed]
57. Mahendran, G.; Rahman, L.U. Ethnomedicinal, phytochemical and pharmacological updates on Peppermint (Mentha × piperita
L.)—A review. Phytother. Res. 2020, 34, 2088–2139. [CrossRef] [PubMed]
58. Bhardwaj, R.L.; Nandal, U. Nutritional and therapeutic potential of bael (Aegle marmelos Corr.) fruit juice: A review. Nutr. Food Sci.
2015, 45, 895–919. [CrossRef]
59. Baliga, M.S.; Bhat, H.P.; Joseph, N.; Fazal, F. Phytochemistry and medicinal uses of the bael fruit (Aegle marmelos Correa): A
concise review. Food Res. Int. 2011, 44, 1768–1775. [CrossRef]
60. Bhattacharyya, P.; Bishayee, A. Ocimum sanctum Linn. (Tulsi): An ethnomedicinal plant for the prevention and treatment of cancer.
Anti-Cancer Drugs 2013, 24, 659–666. [CrossRef]
61. Kumar, A.; Rahal, A.; Chakraborty, S.; Tiwari, R.; Latheef, S.K.; Dhama, K. Ocimum sanctum (Tulsi): A miracle herb and boon to
medical science—A Review. Int. J. Agron. Plant Prod. 2013, 4, 1580–1589.