<![CDATA[Deep learning is widely used to support precise visual analysis. This study aims to analyze and compare the performance of potato leaf image segmentation models using the standard U-Net architecture and U-Net with EfficientNet-B0 through a transfer learning approach. The dataset used consists of 3000 potato leaf images along with segmentation masks obtained from secondary data sources with a division of 1000 per class, then divided into training, validation, and testing data. The research stages include data preprocessing and augmentation, segmentation modeling, model training, and model performance evaluation. Evaluation is carried out using intersection metrics such as Intersection over Union (IoU), mean IoU (mIoU), and Dice Coefficient, with a confusion matrix as an additional analysis tool. The implementation of this study uses the Python programming language supported by the TensorFlow and Keras frameworks, and is run on the Google Colab environment. The results of this study are expected to show the performance differences between the standard U-Net and U-Net with EfficientNet-B0, thereby providing an overview of a more optimal segmentation model for potato leaf image analysis.]]>
Copyrights © 2026
