| Abstrakt: |
Plant diseases can significantly affect crop productivity if not effectively managed. Accurate disease identification is critical for disease control and yield enhancement. Addressing these concerns, the potential application of deep learning techniques for plant disease identification is promising in Indonesia. This research aims to formulate a deep-learning model to detect strawberry (Fragaria sp.) plant diseases. The study encompasses several key phases, including (1) collecting datasets, (2) preprocessing datasets, (3) annotating datasets, (4) configuring and training deep learning models, and (5) validating and evaluating the model. The developed model employs YOLOv7 and YOLOv7-X algorithms, utilizing a dataset of 7337 instances across three disease categories: tip burn, leaf scorch, and anthracnose. These datasets were obtained from publicly accessible repositories. The evaluation of the deep learning model's performance in detecting plant diseases involved using 717 in-field plant images. The outcomes of the evaluation, employing YOLOv7 and YOLOv7-X algorithms, demonstrated accuracy rates of 92.5% and 92.3%, precision levels of 94.5% and 95.1%, and recall values of 90.5% and 89.6%, respectively. These results emphasize the effectiveness of the deep learning model in accurately and precisely identifying diseases in strawberry plants, and it could lay the groundwork for further studies using deep learning to detect diseases in various crops. [ABSTRACT FROM AUTHOR] |
