<![CDATA[The rapid development of deep learning has revolutionized plant disease detection, particularly in precision agriculture. This study aims to compare the performance of a custom Convolutional Neural Network (CNN) and MobileNetV2 in classifying corn leaf diseases into three categories: blight, rust, and healthy leaves. The dataset consists of 303 images captured directly from cornfields in Indonesia, divided into training, validation, and test sets with a 70:15:15 ratio. To overcome data scarcity, data augmentation techniques such as rotation, zoom, and flipping were applied. The custom CNN model and MobileNetV2 (fine-tuned from ImageNet weights) were trained using TensorFlow on Google Colab with a T4 GPU. Experimental results show that MobileNetV2 outperformed the custom CNN in accuracy, precision, recall, and F1-score, demonstrating its efficiency and adaptability for small agricultural datasets. The findings confirm that transfer learning and data augmentation significantly improve classification performance, making MobileNetV2 a lightweight yet accurate solution for corn leaf disease detection in real-world agricultural applications.]]>
Copyrights © 2025
