Optimasi Data Preprocessing dan Hyperparameter Tuning pada Klasifikasi Penyakit Daun Apel menggunakan DenseNet169

Apples are one of the horticultural commodities in Indonesia with production reaching 5,235,955 quintals in 2022, but decreasing to 3,925,628 quintals in 2023. One of the causes of this decline is diseases in apple plants that occur on the leaves, such as scab, black rot, and cedar rust, which can result in a decrease in the quality and quantity of production. Therefore, technology is needed for fast and accurate classification of diseases on apple leaves. This study uses a CNN model with DenseNet169 with optimization on data preprocessing and hyperparameter tuning to improve the accuracy of the apple leaf disease classification model. A total of 36 combinations of data preprocessing and hyperparameter tuning scenarios were tested on the apple leaf image dataset consisting of 4 classes: scab, black rot, cedar rust, and healthy. The optimal scenario is obtained from a combination of RGB + CLAHE with RMSprop optimizer and a learning rate of 0.0001 (P6 + H4), which results in 99.39% accuracy, 99.4% precision, 99.39% recall, and 99.39% f1-score. The results of this study show that the selection of the right preprocessing data and hyperparameter tuning greatly affects the performance of the apple leaf disease classification model.