International Research Journal of Engineering and Technology (IRJET)
Volume: 07 Issue: 07 | July 2020 www.irjet.net
Leaf Classification for Plant Recognition with Deep Transfer Learning
Nandini G1, Roopashree S2, Sai Sharanya3, Shilpa Saboo4, Sandeep Kumar5
1,2,3,4B.E, Dept. Of Computer science & Engineering, Cambridge Institute of Technology, Karnataka, India
5Associate Professor, Dept. Of Computer science & Engineering, Cambridge Institute of Technology, Karnataka,
India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Plant recognition systems that developed by
computer vision researchers, help botanists in faster
recognition and detection of unknown plant species. Until now,
multiple studies focused on the process or algorithms that
maximize use of botanical datasets for plants prediction
modeling, but this method depends on leaf characteristics
which can be changed with botanical data and different
feature extraction techniques. On the other hand, recently, due
to the popularity and successful implementation of deep
learning-based methods in various areas such as image
classification, object detection, and speech recognition, the
researchers directed from traditional feature based methods
to deep learning. In this research, one more efficient method
presented that use transfer learning to recognize plant for leaf
classification, which first uses a pre-trained deep neural
network model for learning useful leaf characteristics directly
from the input data representation. Then use a Multiclass SVM
classifier for leaf classification. It is seen that transfer learning
from a large dataset to limited botanical dataset in plant
recognition task is well done. The proposed method is
evaluated on two well-known botanical datasets and has
succeeded in achieving an accuracy of more than 91%
respectively. The results show that despite the large change in
the number of classes in these two datasets, the proposed
method, have a good performance and show the better result
than methods based on the hand-crafted feature and other
methods based on the deep learning in terms of memory and
precision.
Key Words: Image acquisition, Image Pre-processing, GLCM
Feature extraction, Multiclass SVM classifier.
1. INTRODUCTION
A good understanding of plants for help to recognize new
and rare plant species is essential for improving
pharmaceutical industries, the eco-system balance, agronomy
utilization and robustness. Botanists use variations on the
leaf characteristics as a comparative tool for their study about
plants. On the other hand, due to the progress of science and
genetic topics, many of the compound plants cannot
recognize by ordinary people or even expert people in this
field. For this reason, tried to use computer vision or machine
learning techniques to solve this issue. In computer vision,
despite the many efforts that have been conducted, plant
© 2020, IRJET | Impact Factor value: 7.529
e-ISSN: 2395-0056
p-ISSN: 2395-0072
recognition seems as a challenging and unsolved issue. This is
because a plant in nature has a very similar shape and color.
In addition to plants leaf morphological variations such as
changes in the size, texture, shape, venation, and so on, has
been added to this challenging issue.
Deep learning is a new and modern method for image
processing and data analysis that show high potential and
promising results. Deep learning has been successful in
various fields and recently entered in the field of agronomy
and food products. Therefore, leaf recognition based on deep
learning methods has attracted more attention.
2. LITERATURE SURVEY
In [1], Introduced a novel approach for classification of
plants which was based on the characterization of texture
properties. They have utilized a combined classifier learning
vector quantization along with the radial basis function. The
proposed systems ability to classify and recognize a plant
from a small part of the leaf is its advantageous thing. This
system is mostly applicable as the combined classifier
method produced high performance far superior to other
tested methods as its correct recognition rate was 98.7%
which has been revealed in the result.
In [2], Proposed a method that incorporates shape, vein,
color, and texture features. They have used probabilistic
neural networks (PNN) as a classifier for the plant leaf
classification. Commonly several methods are there for
plant leaf classification but none of them have captured color
information, because color was not recognized as an
important aspect to the identification.
In [3], Proposed a feature fusion technique using the Gabor
filter in the frequency domain and fusing the obtained
features with edge based feature extraction. The extracted
features were trained using 10 fold cross validation and
tested with CART and RBF classifiers to measure its
accuracy.
In [4], This research paper introduced an approach that
combines relatively simple methods which used shape and
texture features. The shape-based method extracts the
contour signature from every leaf and then calculates the
dissimilarities between them. The orientations of edge
| ISO 9001:2008 Certified Journal | Page 1
 International Research Journal of Engineering and Technology (IRJET)
Volume: 07 Issue: 07 | July 2020 www.irjet.net
gradients are used to analyze the macro-texture of the leaf.
The results of these methods are then combined with the
help of incremental classification algorithm which provides
81.1%accuracy.
In [5], This paper concentrates mainly on the procedure of
how the test feature can be compared with the database. The
database is created with different kinds of leaf and other
purpose leaves. For the execution of test leaf a test image is
captured and the parameter is tested. With this the leaf
which gives the closest match will be labeled.
In [6], Current study used the image processing techniques
in order to classification of plants based on leaves
recognition. Two methods called the Gray-Level Co-
occurrence matrix (GLCM) and Principal Component
Analysis (PCA) algorithms have been applied to extract the
leaves texture features. To classify 13 kinds of plants with 65
new or deformed leaves as test images, the Algorithms are
trained by 390 leaves. The findings indicate thatthe accuracy
of PCA method with 98% come out to be more efficiency
compare to the GLCM method with 78% accuracy.
In [7], The important aspect is to develop a system which
classifies the plants. In this paper a novel framework for
recognizing and identifying plants is been proposed. In this
paper leaves can be recognized based on certain methods.
The recognition of the leaves is done based on feature
extraction method and the Zernike moments. Shape, vein,
color and texture features have been used to identify the leaf
and neural network approach is used to classify them and it
provides accurate results in less time.
In [8], This paper proposes a simple and efficient
methodology for Ayurvedic plant classification using digital
image processing and machine vision technology. The three
major phases in proposed methodology are pre-processing,
feature extraction and classification. Pre-processing is done
in order to reduce unwanted noise from the input image. In
feature extraction phase, different morphologic features
such as mean, standard deviation, entropy are extracted
from the pre-processed leaf image. In the third phase, to
classify and recognize plant species. An accuracy of 93.75%
is obtained for the proposed methodology.
4. EXISTING SYSTEM
Deep learning is a new and modern method for image
processing and data analysis that show high potential and
promising results. Deep learning has been successful in
various fields and recently entered in the field of agronomy
and food products. Therefore, leaf recognition based on deep
learning methods has attracted more attention.
In any way, the training of a new deep learning model
from the beginning requires a lot of data, cost, more
computing resources and hours, in some cases, days of
training. On the other hand, in real applications, collecting
© 2020, IRJET | Impact Factor value: 7.529
e-ISSN: 2395-0056
p-ISSN: 2395-0072
and labeling large amounts of data in a particular domain is
costly and time consuming issues, and may even be
impossible. Therefore, in many cases, the idea of collecting a
large number of data in a particular domain may not be
acceptable. This problem poses a challenge and makes it
difficult to use deep learning models.
4. PROPOSED METHODOLOGY
This approach can be a case study to provide improved
accuracy in identification of proper plant leaves. It will a
great help for horticulture and botanical industry as this
framework can be used as an automated tool that can
classify or identify the proper leaf. So that this work can
provide an helping hand to the skilled farmers in identifying
the plant categories
4.1 Image Acquisition
The first step of the system is image acquisition. The image
of a leaf of the plant to be recognized is captured and fed to
the system. Since our system is currently a desktop app, we
have to locate the image via the user interface.
4.2 Image Pre-processing
Pre-processing is used to get the outer shape of the image
from the colored image of the leaf and to remove any kind of
external noises present in an image. The main idea of
preprocessing is to enhance the image details so that
features are clearly found for further processing.
Fig -1: Image Pre-processing Stage
4.3 GLCM Feature Extraction
Features extraction is the next important step after pre-
processing to perform. GLCM feature extraction aims at the
extraction of the relevant information that characterizes
different classes.
The texture features are extracted using GLCM (Gray level co-
occurrence matrix). GLCM creates a gray-level co- occurrence
matrix from particular image. The GLCM involves several
properties and parameters for extracting the texture features.
Properties of GLCM:
⦁Contrast Returns a measure of the intensity contrast
between a pixel and its neighbour over the whole image.
Contrast is 0 for a constant image.
| ISO 9001:2008 Certified Journal | Page 2
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 07 | July 2020 www.irjet.net p-ISSN: 2395-0072

⦁Correlation Returns a measure of how correlated a pixel is 5.2 Sequence Diagram

to its neighbor over the whole image. Correlation is 1 or -1 for
a perfectly positively or negatively correlated image. A sequence diagram is an interaction diagram that shows
Correlation is NaN for a constant image. how objects operate with one another and in what order.
It is a construct of a message sequence chart.
⦁Energy Returns the sum of squared elements in the GLCM.
Energy is 1 for a constant image.

⦁Homogeneity Returns a value that measures the closeness

of the distribution of elements in the GLCM to the GLCM
diagonal. Homogeneity is 1 for a diagonal GLCM.

4.4 Multiclass SVM classifier

This is the last step of detection in which classifier is used to
recognize the kind of the leaf. Classification deals with
matching the given data vectors with one of the trained data
of different classes. Image classification analyzes the
numerical properties of various image features and
organizes data into categories. Classification algorithms
typically employ two phases of processing: training and
testing. In the initial training phase, characteristic properties Fig -3: Sequence Diagram for Leaf Detection
of typical image features are isolated and, based on these, a
unique description of each classification category, i.e.
training class, is created. In the subsequent testing phase, 6. RESULTS AND DISCUSSION
these feature-space partitions are used to classify image
features. 6.1 Dataset
The dataset used in our experiment is collected by our self.
5. SYSTEM DESIGN We pluck the leaf from the plant in the fields near our
System Design is a process of defining the architecture, campus, which consists of intact and fresh leaf images in
modules, interface and the data for the system to satisfy different rotation for 12 plant species class and constructed
by our self. Feature values are extracted from standard
these specific requirements. System Design could be seem as
datasets and they are trained and tested.
a application of a system theory to product development.

5.1 System Architecture

Input Leaf

Features Extract Train Datasets

Leaf

Feature extract
and classification

Classification

Train Database

Multiclass SVM Classifier

Classified results

Fig -2: Proposed System Architecture Fig -4: The sample dataset of leaf images

© 2020, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 3
 International Research Journal of Engineering and Technology (IRJET)
Volume: 07 Issue: 07 | July 2020 www.irjet.net
6.2 Select the leaf image from Test Dataset
Fig -5: Loading the leaf image
6.3 Displaying the result with the accuracy graph
Fig -6: Accuracy Graph
7. CONCLUSION
In this project the main aim is to identify the type of leaf. The
database consists of a set of pre-defined leaf images which
are then compared with the input leaf image to get the
closest match. The leaf classes are identified with the
features extracted using certain algorithms. The given input
leaf is compared with the database created to get the closest
match and the class name of the closest match is being
assigned to the input leaf. The structure analysis is done with
the morphological processing which is nothing but the
structuring elements of binary images. The features are
extracted with the GLCM technique. The classes of leaves are
identified with the features determined. GLCM gives better
structure approximations and hence makes classification
easier. It is found that classification accuracy is better with
SVM classifier than neural network classifier.
© 2020, IRJET | Impact Factor value: 7.529
e-ISSN: 2395-0056
p-ISSN: 2395-0072
8. FUTURE SCOPE
Next step in the future work, different classifiers will be
tested based on our dataset and the results will be recorded.
Here in this project we are considering texture feature.
However, we may have to consider images that contain many
leaves in order to test the ability of the classifiers and we are
looking to consider leaf margins, veins for better
identification of confusing leaves. We can also build a mobile
application of the proposed system, using this one can easily
and efficiently classify the leaves.
ACKNOWLEDGEMENT
We are thankful for our project guide, Prof. Sandeep Kumar
who supported us during the project completion and also
helped us to improve the manuscripts significantly.
REFERENCES
[1] M. Z. Rashad, B. S. el-Desouky,and Manal S. Khawasik,
“Plants Images Classification Based on Textural Features
using Combined Classifier”, International Journal of
Computer Science & Information Technology (IJCSIT), Vol 3,
No. 4, August 2011,pp.93-100.
[2] Abdul Kadir, Lukito Edi Nugroho, and Paulus Insap
Santosa, “Leaf classification using shape, color, and texture”,
International Journal of Computer Trends & Technology
(IJCTT), July-August 2011,pp.225-230.
[3] C. S. Sumathi and A. V. Senthil Kumar, “Edge and Texture
Fusion for Plant Leaf Classification”, International Journal of
Computer Science and Telecommunications, Vol 3, Issue 6,
June 2012,pp. 6-9.
[4] T. Beghin, J. S. Cope, P. Remagnino, & S. Barman, “Shape
and texture based plant leaf classification”, Advanced
Concepts for Intelligent Vision Systems (ACVIS),Vol 6475,
2010,pp.45-353.
| ISO 9001:2008 Certified Journal | Page 4
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 07 | July 2020 www.irjet.net p-ISSN: 2395-0072

© 2020, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 5