A Method for Detecting Citrus Leaf Diseases Improved Based on YOLOv10

Owing to the rapid advancement of deep learning, its application in the identification of crop diseases has emerged as a current research hotspot. Compared with traditional machine learning, its recognition accuracy is typically higher, enabling rapid and accurate identification of pests and diseases. In this paper, based on the YOLOv10 benchmark model, by introducing the CBAM attention mechanism and connecting it after the C2f module, the attention module C2f-CBAM is constructed. Meanwhile, the deformable convolutional network DCNv3 is employed to replace the convolutional network in the SPPF layer, thereby constructing the SPPFv3 module. Experiments indicate that the two improvements respectively have varying degrees of enhancement in accuracy, recall rate, and mAP@50, and the model that integrates both improvements concurrently achieves the optimal effect.