| Autor: |
Subekti, Aulia Haritsuddin K. M., Alfin, Muhammad R., Pinem, Josua G., Pramesti, R. P., Supriyadi, M. R., Chasanah, Umi, Putra, Gilang M., Maulana, Bayu R., Wibowo, Mukti, Budiarti, Dewi H., Muliadi, Jemie, Nugroho, Anto S., Hidayati, D. N., Oktaviani, Avi N., Waluyo, Danang, Besari, Ariza Yandwiputra |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2024, Vol. 3115 Issue 1, p1-9, 9p |
| Abstrakt: |
In Indonesia, there are 2,273 species of fungi that are classified into various genera. With these numerous varieties, fungi bring a lot of impacts on human life and the environment. Fungi can be employed to make foods, create drugs, and make fertile soil. In reverse, fungi also bring many negative impacts, such as causing diseases in humans, plants, and animals. To get the benefit and avoid the harm of the fungi, it needs to be classified in the right group. One way to easily identify and classify fungi is by observing their morphological traits in their image. To accelerate the image-based fungi genera classification process, a multistage convolutional neural network is proposed in this research. There are two schemes in two stages of training, with adjustments in each stage. The result indicates that the scheme with cross-validation in the first step gives better classification result and can handle overfitting issues. Scheme with cross-validation gives a higher average f-1 score of 59 % compared to the scheme without cross-validation with an average f-1 score of 28%. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
