An Improved Convolutional Neural Network Algorithm and Its Application in Multilabel Image Labeling

Figure 2

Structure of a two-channel convolutional neural network. Conv = convolutional layer; Pool = pooling layer; Fc = fully connected layer. The x in conv-x denotes the number of convolution kernels in the layer; the size of the convolution kernel or pooling window in the layer is y × y; the z in Fc-z denotes the number of neurons in the fully connected layer; and dropout shows the addition of a dropout layer to the original layer.