<![CDATA[Bananas are an important commodity to support the Indonesian economy, but the frequent occurrence of banana spot disease in tropical countries can cause problems for this industry. The purpose of this study is to build and compare two models, namely ResNet50, which is often used as the main comparison, with EfficientNetV2M as a newer model. The dataset was obtained from two sources on the Mendeley data platform. The combined dataset consists of 1938 data labeled with 5 classes: cordana, pestalotiopsis, black sigatoka, yellow sigatoka, and healthy. The preprocessing process was carried out by dividing the data into training, validation, and testing with a ratio of 70:20:10, then resizing and assigning weights to each class. Two models were trained with the same parameters and evaluation metrics such as accuracy, precision, recall, f1-score, and loss. The test evaluation results show the results of the EfficientNetV2M model test with an accuracy metric value of 94,92%, precision of 95,49%, recall of 92,23%, f1-score of 93,70% and loss of 17,96%. While ResNet50 with an accuracy value of 91,88%, precision of 90,97%, recall of 87,59%, f1-score of 89,14%, and Loss of 20,87%. Based on these evaluation results, EfficientNetV2M is a model with superior performance for the classification of banana leaf spot disease. This research is expected to be a reference for determining an effective model and developing a classification system for banana leaf spot disease.]]>
Copyrights © 2025
