Analysis and Factors Affecting Accuracy

Oct 13, 2025 by pynkocean.com 129 views

Introduction

Herbal plants from the Genus Curcuma, such as turmeric and ginger, have many species with very similar characteristics. The similarity of these features often makes it difficult to distinguish, especially in the pharmaceutical field where errors in identification can have a serious impact. Plant leaves are the most often used in making herbal medicines, so research on the leaves and organelles, such as stomata, becomes very important. Accurate identification of plant species is crucial in the pharmaceutical industry, where the wrong identification can lead to serious consequences.

To overcome this challenge, an effective approach is needed in identifying the types of organelles found in the leaves of herbal plants. One promising method is the application of artificial neural networks, especially through the Deep Learning technique with the Convolutional Neural Network (CNN). This study aims to classify the microscopic image of stomata of two plant species in the Curcuma genus, namely turmeric and ginger. Deep Learning has shown great potential in image classification tasks, and this study aims to explore its application in the classification of microscopic images of herbal stomata.

Research Methodology

In this study, the Deep Learning implementation was carried out using CNN which was supported by the Gabor filter process and feature extraction using Gray Level Cooccurrency Matrix (GLCM). Gabor filters are effective in capturing frequency and orientation information in the image, while GLCM helps in image texture analysis. In this way, the system is expected to produce an accurate model in classification. The use of Gabor filters and GLCM is a key aspect of this study, as it allows for the extraction of relevant features from the microscopic images.

A total of 160 microscopic images of Curcuma herbal plants are used as training data. The resulting model reached training accuracy of 93.1%, showing excellent capabilities in recognizing relevant image features. Meanwhile, in the test data consisting of 40 images, the accuracy achieved is 92.5%. These results indicate that the model built can be relied upon in identifying the stomata of different plant species. The high accuracy of the model is a significant finding, as it suggests that Deep Learning can be a useful tool in the classification of microscopic images of herbal stomata.

Analysis and Factors Affecting Accuracy

From the results of system testing, it was found that classification accuracy was influenced by several key factors:

1. Early Image Form

Quality and initial image resolution greatly affects the final model. A clearer and more detailed image tends to provide better information for nerve networks, thereby increasing accuracy. The quality of the input image is a critical factor, as it can affect the performance of the model.

2. CNN Parameters

Adjusting parameters in CNN, such as the number of layers, kernel sizes, and activation functions, play an important role in how the learning model of data. Appropriate adjustments can help the model in generalizing features better. The choice of CNN parameters is a crucial aspect of this study, as it can affect the performance of the model.

3. Number of Dataset

Availability of a large and varied dataset is also important to improve the ability of the model in identifying fine differences between species. More training data will usually increase classification accuracy. The size and diversity of the dataset are critical factors, as they can affect the performance of the model.

Conclusion

Deep learning implementation using Convolutional Neural Network shows great potential in the classification of microscopic image of herbal stomata from the genus curcuma. With a high level of accuracy in training data and test data, this method can be a useful tool in the pharmaceutical field to help identify plant species. Further research with dataset variations and parameter adjustments can increase the reliability and efficiency of this classification system. Innovations in this method are not only beneficial for the academic world, but also have important implications in the herbal and pharmaceutical industries.

Future Work

This study has shown the potential of Deep Learning in the classification of microscopic images of herbal stomata. However, there are several areas that require further research:

Improving the accuracy of the model: Further research is needed to improve the accuracy of the model, particularly in the test data.
Increasing the size and diversity of the dataset: A larger and more diverse dataset is needed to improve the performance of the model.
Exploring other Deep Learning architectures: Other Deep Learning architectures, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks, may be explored to improve the performance of the model.

By addressing these areas, this study can contribute to the development of a more accurate and reliable classification system for microscopic images of herbal stomata.

Q1: What is the main objective of this study?

A1: The main objective of this study is to classify the microscopic image of stomata of two plant species in the Curcuma genus, namely turmeric and ginger, using Deep Learning implementation with Convolutional Neural Network (CNN).

Q2: What is the significance of this study?

A2: This study is significant because it explores the application of Deep Learning in the classification of microscopic images of herbal stomata, which is a crucial task in the pharmaceutical industry. Accurate identification of plant species is essential to ensure the quality and safety of herbal medicines.

Q3: What are the key factors that affect the accuracy of the model?

A3: The key factors that affect the accuracy of the model are:

Early image form: Quality and initial image resolution greatly affects the final model.
CNN parameters: Adjusting parameters in CNN, such as the number of layers, kernel sizes, and activation functions, play an important role in how the learning model of data.
Number of dataset: Availability of a large and varied dataset is also important to improve the ability of the model in identifying fine differences between species.

Q4: What is the role of Gabor filters and GLCM in this study?

A4: Gabor filters are effective in capturing frequency and orientation information in the image, while GLCM helps in image texture analysis. These features are used to extract relevant information from the microscopic images, which is then used to train the CNN model.

Q5: What are the limitations of this study?

A5: The limitations of this study are:

Small dataset size: The dataset used in this study is relatively small, which may limit the generalizability of the results.
Limited CNN architectures: This study only explores the use of CNN architecture, and other Deep Learning architectures may be more suitable for this task.
Limited parameter tuning: This study only explores a limited set of CNN parameters, and further research is needed to explore other parameters and their effects on the model.

Q6: What are the future directions of this research?

A6: The future directions of this research are:

Improving the accuracy of the model: Further research is needed to improve the accuracy of the model, particularly in the test data.
Increasing the size and diversity of the dataset: A larger and more diverse dataset is needed to improve the performance of the model.
Exploring other Deep Learning architectures: Other Deep Learning architectures, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks, may be explored to improve the performance of the model.

Q7: What are the potential applications of this research?

A7: The potential applications of this research are:

Pharmaceutical industry: This research can be used to develop a more accurate and reliable classification system for microscopic images of herbal stomata, which can be used to ensure the quality and safety of herbal medicines.
Herbal industry: This research can be used to develop a more accurate and reliable classification system for microscopic images of herbal stomata, which can be used to identify and classify different herbal species.
Academic research: This research can be used to explore the application of Deep Learning in the classification of microscopic images of herbal stomata, which can contribute to the development of new methods and techniques in this field.