Research on A Citrus Leaf Disease Detection Model Based on Yolov5-CBF

In order to achieve automatic detection of citrus leaf diseases and meet the high accuracy recognition requirements, this study proposes a citrus leaf disease detection model named Yolov5-CBF based on Yolov5. Firstly, to enhance the model's ability to capture characteristic information of small diseases, a coordinate attention mechanism was incorporated into the model to improve the detection accuracy of such diseases. Secondly, to address the impact of varying disease characteristics, the original PANet architecture was replaced with BiFPN in order to improve the feature extraction ability of the model on citrus disease information. Finally, to mitigate the impact of citrus leaf background, the loss function was improved into Focal loss to improve the overall detection accuracy. Compared to Yolov5, the improved YOLOV5-CBF exhibits higher detection accuracy with an increased mAP value of 5.85%. The enhanced Yolov5-CBF was applied to the disease detection of citrus leaves, which could effectively improve the detection accuracy and improve the average accuracy of detection.