GSW-YOLOv7:一种基于改进YOLOv7的玉米叶病害检测方法

GSW-YOLOv7:A Maize Leaf Disease Detection Method Based on Improved YOLOv7

准确识别农田中常见类型的玉米叶病害并及时治疗是提高玉米产量的关键,可以提高效率并降低种植成本。为了解决玉米叶病害识别精度不高、检测效率低以及在移动端设备难以部署的难题,本文提出了一种改进的名为GSW-YOLOv7的目标检测模型。首先,设计了GS-ELAN结构优化模型颈部,降低模型的参数量。其次,将简单且无需参数的SimAM注意力机制融入网络结构中,在不增加额外参数的情况下提高检测的精度,获取更具代表性的玉米叶病害特征。最后,采用高效的Wise-IoU损失函数以加快收敛速度,并提高模型的精度。实验结果表明,GSW-YOLOv7网络模型平均精度(mAP)为85.60%,检测速度为26.79FPS,该模型在检测任务中权衡了模型的检测精度和检测速度,与YOLOv5、YOLOX等算法相比,该算法性能最好,能够快速、准确地检测常见的玉米叶病害,为农业生产提供了创新的解决方案。

Accurately identifying various types of maize leaf diseases in the field and treating them in a timely manner is key to increasing maize yield,improving efficiency,and reducing planting costs.To address the issues of low accuracy in identifying maize leaf diseases,low detection efficiency,and difficulties in deployment on mobile devices,we propose an improved tar-get detection model named GSW-YOLOv7.First,the GS-ELAN structured optimization model is proposed to optimize the model's neck,reducing the parameter count.Second,the SimAM attention mechanism is incorporated into the neck of the YOLOv7 structure with the aim to improve recognition accuracy without increasing additional parameters,thereby obtaining more representative features of maize leaf diseases.Finally,an efficient Wise-IoU loss function is applied,which improves the accuracy of the model.The experimental results show that the GSW-YOLOv7 network model achieved a mean average precision(mAP)of 85.60%and a detection speed of 26.79 frames per second(FPS),which balanced detection accuracy and speed in detection tasks.The algorithm has the best performance in comparison with algorithms such as YOLOv5 and YOLOX,which are capable of detecting common maize leaf diseases quickly and accurately.Thus,the model provides an innovative solution for agricultural production.