Classification and Recognition of Succulent Plant Images Based on Improved MobileNetV3
In order to complete the high-precision classification and identification of succulents and deepen the cultivation field of succulents,10 kinds of succulents were selected as data categories,based on the morphological characteristics of their roots,stems and leaves,and comparative tests were added.The same method was used to train four deep convolutional models,including AlexNet,VGG16,ResNet50 and MobileNetV3,without transfer learning,with a learning rate of 0.001 specified by hyperparameter and Adam optimizer.The results show that MobileNetV3 has the best overall learning effect.On the basis of transfer learning,Mo-bileNetV3 model continues to be improved,and cavity convolution and RAdam optimization algorithm are introduced into the con-volution layer for parameter optimization.The average test recognition accuracy rate can reach 99.7%.The improved MobileNetV3 network model has a good effect on the identification of succulents.
NETL
NSTL
万方数据
