Fruit Classification Model Based on Improved Darknet53 Convolutional Neural Network
Hui Wang, Fan Zhang, Li Wang
Abstract
In order to solve the problem of fruit classification under complex background, this paper proposes a fruit classification model based on improved darknet53 convolutional neural network. In this model, the group normalization method is used instead of the original batch normalization method, and the softmax classifier with 22 tags is used instead of the softmax classifier in the original darknet53. These changes can optimize the model structure and parameters. The experimental results show that this method can effectively extract the multi-layer features of fruit image, and does not depend on the batch size, and obtain higher accuracy than the original model. Compared with darknet-53, inception V3 and vgg-16, this mode has the best performance with an accuracy of 95.6%.