基于改进MobileNetV3的轻量化枇杷病害识别

Lightweight Loquat Disease Identification Based on Improved MobileNetV3

为了枇杷病害能够及时检测,提高枇杷病害识别的准确率,在MobileNetV3基础上提出一种轻量化枇杷病害识别算法。首先,用PConv替换MobileNetV3中的DWConv,设计新的block结构;然后,引入CBAM注意力模块,从通道和空间上增加特征表达能力以提高模型准确率;最后,重新设计网络结构得到改进的MobileNetV3模型。实验表明,改进后的算法准确率达97.79%,模型参数量为1.14M,检测速度为21.9fps。该算法实现轻量化效果,可快速准确地对枇杷病害进行识别,为移动端实现枇杷病害识别提供新的技术支持。

In order to detect loquat diseases in time and further improve the accuracy of loquat disease recognition,a lightweight loquat disease recognition algorithm is proposed on the basis of MobileNetV3.Firstly,PConv is used to replace DWConv in MobileNetV3 network to design a new block structure.Then the CBAM attention module is introduced to increase the feature expression ability from channel and space improve the model accuracy.Finally,the network structure is redesigned to obtain an improved MobileNetV3 model.Experiments show that the accuracy of the improved algorithm is 97.79%,the model parameters is 1.14M,and the detection speed is 21.9 fps.This algorithm achieves the lightweight effect,which can quickly and accurately identify loquat diseases,and provides new technical support for the mobile implementation of loquat disease identification.