Klasifikasi Varietas Tanaman Alpukat Berdasarkan Citra Daun Menggunakan Metode Convolutional Neural Network (CNN)

Abstract

The identification of avocado plant varieties is commonly performed through manual observation of leaf morphology, which requires specofoc expertise and a relatively long time. This process may also lead to misidentification because several avocado varieties have highly similar leaf shapes and textures. Therefore, an automated method is needs to identify avocado varieties quickly and accurately. This study focuses on evaluating the performance of a Convolutional Neural Network (CNN) in classifying ten varieties of avocado plants using leaf images. The approach applied in this research utilizes the MobileNetV2 architecture combined with a transfer learning strategy to enhance both training efficiency and overall model performance. The dataset in this study consist of avocado leaf images with variations in leaf maturity, including young, intermediate, and mature stages. Before training was conducted, the dataset was split into training, validation, and testing subsets. Several evaluation metrics were used to measure the model’s performance, including accuracy, precision, recall, f1-score, and confusion matrix analysis. The experimental results indicate that the MobileNetV2-based CNN achieved of 93% on the testing data. Furthermore, the training curves demonstrate a stable convergence between training accuracy and validation accuracy, indicating that the model is capable of generalizing well to unseen data. However, some misclassifications were still observed, particulary among varieties with very similar lear morphological features. Overall, the findings suggest that the CNN approach using the MobileNetV2 architecture is effective based on leaf images.