<![CDATA[The detection and management of sweet potato leaf diseases play a vital role in ensuring sustainable crop yields and reducing agricultural losses. This study proposes an automated segmentation approach using the U-Net convolutional neural network to detect disease regions on sweet potato leaves. The dataset, consisting of leaf images and corresponding masks, underwent a structured preprocessing pipeline including resizing, normalization, and reshaping. The U-Net architecture, comprising an encoder-decoder structure with skip connections, was trained on 70% of the dataset and evaluated using accuracy, Intersection over Union (IoU), and Dice coefficient. Experimental results show that the model achieved an accuracy of 94.6%, IoU of 0.88, and a Dice coefficient of 0.92, indicating strong segmentation performance. Visual comparison between predictions and ground truth masks further confirms the model’s effectiveness in isolating disease regions. This research demonstrates the potential of U-Net as a reliable deep learning framework for plant disease detection and contributes to the development of intelligent agricultural monitoring systems.]]>
