Disease detection of mulberry based on YOLOv8s-GRE model

Mulberry disease seriously affects the growth and development of mulberry, the yield and quality of mulberry leaves and mulberries, and brings huge economic losses to the sericulture industry. Therefore, it is very important to strengthen the identification and monitoring of mulberry disease. A lightweight YOLOv8s-GRE network was designed to efficiently detect mulberry disease under complex background in mulberry orchard. It integrates the lightweight and efficient layer aggregation network GELAN in YOLOv8s to make the model more lightweight and reduce the calculation cost. Fusion receptive field attentional convolution can extract disease image information more effectively and enhance feature extraction ability. In addition, ECA attention mechanism is introduced to help the network locate the disease area more accurately. Inner-SIoU loss function is used to improve the accuracy of disease identification and accelerate the convergence of the network. The experimental results show that the improved model has 5.2M parameters, 20.9G computation, 86.0% accuracy, 85.7% recall and 90.7% mAP. Compared with YOLO v8s, the average accuracy (mAP) of the proposed method is increased by 1.0%, the number of model parameters is reduced by 53.2%, and the calculation amount is reduced by 26.4%. While ensuring the detection accuracy and real-time performance, the size of the model is greatly reduced, which is easy to deploy in embedded devices, and well meets the demand for real-time detection of mulberry disease. This study can provide a reference for detection of mulberry disease.