|
Efficiency of plant diseases classification by convolutional neural network with optimization algorithm and activation function |
|---|---|
| รหัสดีโอไอ | |
| Creator | A-nanya Promkot |
| Title | Efficiency of plant diseases classification by convolutional neural network with optimization algorithm and activation function |
| Publisher | The Association of Council of IT Deans (CITT) |
| Publication Year | 2567 |
| Journal Title | Journal of Information Science and Technology |
| Journal Vol. | 14 |
| Journal No. | 1 |
| Page no. | 34-44 |
| Keyword | Image Classification, Image Processing, Convolution Neural Network, Plant Disease |
| URL Website | https://tci-thaijo.org/index.php/JIST |
| Website title | Journal of Information Science and Technology |
| ISSN | 2651-1053 |
| Abstract | Plant diseases are a problem that has a huge impact on farmers. Detection of plant diseases at an early stage to effectively control the spread of germs Therefore, it plays an important role in the agricultural industry. However, traditional approach requires extensive knowledge of the expert. It is expensive and requires a lot of labor. Nowadays, with the advancement of information technology, machine learning and deep learning has been applied to automatic identification of plant disease. Currently, convolutional neural network methods is a method that has been recognized for its efficiency in image classification. The objective of this research is to find appropriate values for the ResNet50 method with optimization algorithms, including AdaDelta, AdaGrad, Adam, RMSprop, and SGD, and activation functions including ReLU, Sigmoid, and Tanh ,for plant disease classification by using plant leave image. Evaluated performance of plant disease classification by using the PlantVillage dataset. The results showed that the ResNet50 method with RMSprop optimizer and Sigmoid activation function gave the highest Accuracy value of 0.94, Precision value of 0.94, Recall 0.93, and F -measure is equal to 0.93. Therefore, it can be concluded that in selecting a model for learning to achieve the most efficient results should consider additional factors including algorithms to increase the efficiency of the model and stimulation functions |
