Deep Learning-Based Strawberry Disease Detection: An Integrated Approach Using YOLOv8, EfficientNetV2, and DeepLabV3+

Strawberry fruits are highly vulnerable to diseases such as leaf scorch, gray mold, anthracnose, powdery mildew, leaf blight, rust, and verticillium wilt that impact the quality and quantity of the fruits. The current approach of disease detection by humans is slow with many errors, so automated detection is necessary. This paper introduces a deep learning hybrid structure consisting of YOLOv8 for object detection, EfficientNetV2 for classification, and DeepLabV3+ for segmentation is proposed to accomplish precise detection and analysis of seven strawberry plant diseases. The data were acquired from Roboflow as images showing healthy and infected strawberry plants that include their associated labels. Our suggested approach provides the potential for accomplishing precise plant disease localization, classification, and segmentation, enabling improved diagnosis and early treatment. Our results show that the suggested hybrid approach has good performance than traditional approaches and obtains high accuracy in precision, recall, and segmentation. Our research makes a contribution to precision agriculture through the evolution of an AI-driven, flexible, and optimized disease detection system that farmers and agronomists can utilize to manage plant health more effectively. Our paper demonstrates current agricultural demand for deep learning through potential applications and limitations along with proposed future explorations.