Research on the Influence of Convolutional Neural Network Parameters on Fruit Classification

Classification and identification of fruit has significant application prospects in smart agriculture, which plays an integral part in the modernization of agricultural production. This paper constructs a convolutional neural network model applied to the classification and identification of various fruits. Concretely, the study explores how learning rate, batch size and the number of epoch, three parameters of this model, influence the network performance. The experimental results show that when other parameters remain unchanged, errors trained by the network model significantly decrease as the learning rate increases, the batch size reduces and the number of epoch accumulates within a certain range. In addition, the image recognition rate of the test set can also be improved. To be specific, simply through parameter debugging, the errors can be reduced by an order of magnitude and the recognition rate can be improved by 1% to 2%.