Authors

Abstract

In order to achieve efficient and accurate detection of common corn leaf diseases such as leaf blight, gray spot disease, and rust, a corn leaf disease detection method based on CA-YOLOv8 was proposed. In this method, the Coordinate Attention(CA) attention mechanism was added after the feature map output from the Neck part to enhance the feature extraction capability of the model. The experimental results showed that the precision，recall and mean average precision(mAP) of the CA-YOLOv8 model on the test set were 94.08%, 90.53% and 97.38%, respectively. Compared with the YOLOv8, YOLOv8+SE and YOLOv8+CBAM models, the mAP was improved by 2.15, 0.86 and 2.35 percentage points, respectively. Compared with Faster R-CNN, YOLOv5s, YOLOv7, and YOLOv8 models, the mAP has increased by 63.53, 29.24, 3.21, and 2.15 percentage points, respectively. The study showed that the CA-YOLOv8 model can provide a technical reference for the development of a portable intelligent corn leaf disease detection system.

Abstract in Chinese

为了实现以叶枯病、灰斑病、锈病等玉米常见叶病的高效准确检测，提出了一种CA-YOLOv8的玉米叶病检测方法。该方法在 Neck 部分输出的特征图之后加入CA注意力机制，以便提升模型的特征提取能力。试验结果表明，CA-YOLOv8模型在测试集上的的精确率、召回率和平均精度均值分别为94.08%、90.53%和97.38%。对比YOLOv8、YOLOv8+SE和YOLOv8+CBAM模型，平均精度均值mAP分别提升了 2.15、0.86、2.35个百分点。与Faster R-CNN、YOLOv5、YOLOv7和YOLOv8模型相比，mAP分别提升了63.53、29.24、3.21和2.15个百分点。研究表明，CA-YOLOv8模型能够为便携式智能玉米叶病检测系统开发提供技术参考。