<![CDATA[Leaf diseases are one of the primary factors contributing to reduced agricultural productivity, especially in tropical regions such as Indonesia. Early identification of infected areas is crucial to prevent further disease spread and minimize crop losses. This study proposes the application of YOLOv8 instance segmentation with an in-training augmentation strategy to precisely detect and segment leaf disease areas. The dataset consists of 3,528 diseased leaf images, with 588 original images used for validation and 2,940 augmented images for training. The augmentation process includes mosaic, HSV adjustment, spatial transformations, RandAugment, and other techniques to enhance the model’s generalization capability. Model performance was evaluated using mAP, precision, recall, and Intersection over Union (IoU) metrics. Experimental results demonstrate that the model achieved stable performance with high mAP@50, precision and recall approaching optimal values, and an average IoU above 0.5. This approach is effective and has strong potential for implementation in early plant disease detection systems, supporting farmers in making timely and accurate decisions.]]>
Copyrights © 2025
