SFLD-YOLO: A Lightweight Strawberry Fruit and Leaf Disease Detection Method

Strawberry diseases, which predominantly affect the fruits and leaves, are critical factors impacting crop yield. To address the limitations of existing strawberry disease recognition models—namely, excessive parameter count, suboptimal detection accuracy, weak environmental robustness, and insufficient capability in detecting small targets—this study proposes a lightweight detection method named SFLD-YOLO (Strawberry Fruit and Leaf Diseases-YOLO) based on an improved YOLOv11 architecture. First, the network is enhanced by integrating a C3K2-WTConv module based on wavelet transform, which expands the multi-scale receptive field while significantly reducing the computational complexity of the model. Second, an Efficient Multi-Scale Attention mechanism is dually embedded into the neck and head networks. This synergistic deployment minimizes the interference of complex background noise, dynamically redirects the network's focus toward critical lesion regions, and substantially enhances the detection robustness. Finally, to augment the perception capabilities for early-stage and minute disease symptoms, a dedicated 160 × 160 small object detection head is incorporated into the architecture. Experimental results demonstrate that the proposed SFLD-YOLO achieves a mAP, Precision, and Recall of 91.3%, 98.2%, and 97.0% on the customized strawberry disease dataset, respectively. Compared to the YOLOv11 baseline, these metrics reflect absolute improvements of 5.5%, 1.9%, and 3.0%, respectively, while the total parameter count is drastically compressed to merely 2.4 M. Furthermore, visualization analyses validate the model's superior accuracy in bounding box localization, which effectively mitigates missed and false detections. This study provides a highly efficient and robust reference scheme for the lightweight deployment of intelligent disease management in actual agricultural production.