Plant Pest and Disease Detection Based on Improved YOLOv10

With the development of the times, machine learning is increasingly widely applied in the agricultural field, especially in plant disease detection. To address the difficulties in recognizing early-stage tiny lesions, low accuracy, and complex field environments, this paper proposes a plant disease detection model based on YOLOv10. By introducing the Large Kernel Attention (LKA) mechanism and the EIoU loss function, the disease recognition accuracy in complex field scenes is optimized. The LKA module enhances the model's perception ability for scattered tiny lesions by decomposing large convolution kernels, while the EIoU loss function precisely locates the boundaries of disease regions. Experiments cover 30 typical diseases including apple scab, tomato late blight, and grape black rot. The model was trained and validated for effectiveness using more than 1000 field images. Experimental results show that the improved model achieves a precision of 95.9% and a recall rate of 89.7% in the plant disease detection task. Compared to the baseline model, mAP@0.5 and mAP@0.5:0.95 are improved by 4.9% and 3.1%, respectively. This provides an efficient and reliable technical solution for plant disease detection and identification.