Apple Leaf Disease Identification Method Based on Improved YoloV5

In this paper, an improved YoloV5s based identification model, named as GSE_YoloV5s, is proposed for apple leaf disease identification, which can efficiently address the problem of high storage and computational resource consumption for the YoloV5s model. Firstly, the GhostBottleneck module is used to replace the original CSPBottleneck module to reduce the parameters and computation of the YoloV5s; meanwhile, by adding the channel attention module SE (Squeeze-and-Excitation), the model’s detection performance for small target lesions is improved. The experimental results show that, the improved GSE_YoloV5s model can reduce the number of parameters and computational effort by 40%, as compared with the YoloV5s model, and the AP (Average Precision) achieves 83.4%, which can effectively detect apple leaf diseases.