<![CDATA[Mango is one of the important fruits in Indonesia, but its production is often disrupted by leaf diseases and pests that are difficult to detect early. Manual disease recognition methods usually depend on observers and are not always accurate. This study aims to create an automated system to classify mango leaf diseases, using deep learning techniques based on the Convolutional Neural Network (CNN) algorithm. This study also compares three models, namely VGG16, DenseNet121, and InceptionV3, by applying the transfer learning method. The dataset used consists of 4,000 images divided evenly into 8 categories, consisting of 7 types of diseases (Anthracnose, Bacterial Canker, Cutting Weevil, Die Back, Gall Midge, Powdery Mildew, Sooty Mold) and 1 category of healthy plants. Evaluation was carried out using the 5-Fold Cross-Validation method to ensure valid results. The results show that all three models are able to provide an accuracy of more than 90%. The VGG16 model showed the best and most stable performance, with an accuracy of 93.25%, a Precision of 0.93, a Recall of 0.93, an F1-Score of 0.93, and an AUC-ROC of 0.98. Meanwhile, InceptionV3 achieved an accuracy of 92.38% and DenseNet121 reached 91.25%. Therefore, VGG16 is recommended as the primary model due to its better ability to extract texture features and accurately recognize mango leaf diseases. VGG16 architecture is able to outperform complex models in efficiently extracting mango leaf texture features, making it very potential to be used as a basis for real-time plant disease diagnosis applications for farmers]]>
