1. Visvesvaraya Technological University
Belagavi,Karnataka
National Education Society ®
JNN College of Engineering
Department of Computer Science & Engineering
Technical Seminar presentation on
“Herbal leaf Identification Using Deep Learning
Neural Network”
Presented by
Jeevitha V 4JN20CS040
Under the Guidance of :
Mrs. Radhika S K B.E, M. Tech.,
Assistant Professor,
Dept. of CS&E.
Co-ordinator:
Dr. Manjula G R M. Tech.,Ph.D
Professor,
Dept. of CS&E.
3. Herbal plants can be used as an alternative to the natural healing of
diseases. However, identifying them is challenging due to the vast number of
species and their unrecognized existence. Accurate identification of these plants
requires specialized knowledge. To address these challenges, a smart herbal leaf
recognition system is needed to assist in the identification and conservation of
medicinal plants. This study aims to identify and authenticate herbal leaves using
the convolutional neural network and Long Short-Term Memory (CNN-LSTM)
methods. Implementation of this approach can contribute to the identification and
conservation of valuable medicinal plants.
Abstract
4. 1. Introduction
• Herbal plants represent biodiversity and are often used as alternatives to traditional
medicine. These plants have been used throughout the world for centuries to maintain health
and treat various diseases. The existence of herbal plants is currently not widely known by
the public due to many types of medicinal plants so that people find it difficult to
distinguish between the types of herbal plants.
• Manual identification is difficult because it takes time and can be inaccurate at the same
time. This is because the leaf colour is almost the same, and some types of herbal plants
have almost the same texture and shape. As a result, proper medicinal plant identification
system is required.
• To overcome these challenges, a model utilizing Deep Neural Networks has been developed
that can precisely identify medicinal plants with the help of modern computing technology.
5. continued…
• In the Described work, A leaf image is taken as input, which further goes through
pre-processing, feature extraction and classification and as a result, we obtain
scientific name of the plant.
Fig 1 Block diagram of the described work
6. 2. Problem Description
The lack of awareness among people about medicinal plants has led to
the issue of medicinal plant extinction. Obtaining information about their
medicinal properties usually requires prior exposure and can be overwhelming for
those who are not well-versed in this area. Identifying medicinal plants may also
require the assistance of experts. To tackle these challenges, a Deep Neural
Networks-based CNN-LSTM model has been developed to accurately identify
medicinal plants using modern computing technology. This approach can make it
easier for people to access information about medicinal plants and reduce the
negative impact of medicinal plant extinction on the economy
7. 3. Literature Survey
[2]“Automatic plant recognition using convolutional neural network on Malaysian
medicinal herbs: the value of data augmentation”
Authors: Noor Aini Mohd Roslan, Norizan Mat Diah, Zaidah Ibrahim, Yuda Munarko, Agus Eko
Minarno.
Description:
The paper explores automatic plant recognition using a Convolutional Neural
Network (CNN) on Malaysian medicinal herbs. It compares the performance of the CNN
model on real and augmented datasets, demonstrating the value of data augmentation in
improving accuracy.
Advantages:
• Data augmentation techniques, like rotation and flipping, enhance model accuracy.
• The study showcases the effectiveness of CNNs for plant recognition, particularly in the
context of Malaysian medicinal herbs.
Demerits:
• Limited generalizability to other plant species or regions.
• Computational and time costs associated with data augmentation may be significant.
8. 3. Literature Survey continued...
[3]“Medicinal Plant Identifcation in Real-Time Using Deep Learning Mode”
Authors: S. Kavitha, T. Satish Kumar, E. Naresh, Vijay H. Kalmani, Kalyan
Devappa Bamane, Piyush Kumar Pareek
Description:
The research presents a vision-based method for real-time identification
of medicinal plants using a deep learning model. It focuses on six plant species,
utilizing image data from Kaggle and deploying the model on a cloud-based
mobile app.
Advantages:
• Rapid and precise identification of medicinal plants.
• Accessibility through a mobile app and cloud deployment.
Demerits:
• Provides only 85% accuracy.
• Limited dataset consisting of only six varieties of medicinal plants.
9. 3. Literature Survey continued...
[4]“Assessing deep convolutional neural network models and their
comparative performance for automated medicinal plant identification from
leaf images”
Authors: Biplob Dey, Jannatul Ferdous, Romel Ahmed, and Juel Hossain.
Description:
The paper evaluates seven deep learning algorithms for automated
medicinal plant identification using a dataset of 5878 images representing 30
species. It aims to assess algorithm performance in accurately classifying plant
species.
Advantages:
• Offers a detailed assessment of deep learning algorithms for plant identification.
• Uses a diverse dataset of 5878 images to enhance the generalizability of
findings.
Demerits:
• Works for small datasets.
• High Computational cost and time required.
10. 3. Literature Survey continued...
[5]“Identification and classification of medicinal plants using leaf with deep
convolutional neural networks”
Authors: Nilesh S. Bhelkar, Dr. Avinash Sharma
Description:
Utilizing deep convolutional neural networks enhances the precision of
identifying medicinal plants through leaf characteristics. Machine learning
algorithms streamline plant classification processes efficiently.
Advantages:
• Efficient automated plant classification.
• Improved accuracy in identifying medicinal plants.
Demerits:
• Challenges with complex data and interpretability of models.
• Potential reduction in human involvement and traditional knowledge reliance.
11. 4. System Design & Architecture
• The Described work involves the use of a convolutional neural network and
Long Short-Term Memory (LSTM) for the classification of herbal leaves. The
process of classifying leaves involves several phases, including dataset
acquisition, pre-processing, feature extraction, and classification.
Fig 2 Process of medical plant recognition
12. 4. System Design & Architecture
• Before being processed by CNN and LSTM, the herbal leaf image data is
preprocessed to remove noise, clarify image features, reduce image size, and
transform the image data into a format that can be processed by CNN and
LSTM.
Fig 3 Pre-processing of Images.
13. 4. System Design & Architecture
• The herbal leaf image input is processed using CNN. Each image data is convoluted and then
pooled using the max-pooling layer generates a feature map. This is processed via dropout to
get the desired result.
• The results of this LSTM process will be flattened to form a fully connected layer. For the
classification, the Softmax function is used.
Fig 4 CNN-LSTM Architecture.
14. From the results of the standardized training data, it is evident that the accuracy obtained using
the CNN-LSTM model is higher compared to the accuracy achieved with the CNN model.
5. Results
Fig 5 Accuracy training data with
CNN model.
Fig 6 Accuracy training data with
CNN-LSTM model.
15. 5. Results
• The training process was divided into two stages, involving the training of data with
the CNN model and the CNN-LSTM model separately.
Fig 7 Loss training data with
CNN model.
Fig 8 Loss training data with
CNN-LSTM model.
16. The lack of knowledge about medicinal plants poses a challenge in identifying
the correct plant, which can ultimately lead to the extinction of these valuable resources.
However, modern technology has provided a solution through the development of a
Deep Neural Network model using CNN-LSTM that accurately identifies medicinal
plants. In this work, a leaf image is taken as input, which further goes through pre-
processing, feature extraction, and classification, resulting in obtaining the scientific
name of the plant. With the assistance of this model, individuals can accurately identify
medicinal plants, leading to the creation of more effective treatments and health
remedies.
6. Conclusion