FIGURE 1.

The structures of three developed deep neural networks. The orange, brown, and red cubes represented the V block, R block, and I block in Figure 2, respectively. The dark green and light green cubes represented the max-pooling layer and the average pooling layer, respectively. The yellow and blue cubes represented a one-dimensional convolutional layer, BN cascade activation function ELU, respectively. The striped bars represent flattened layers, and the light blue bars represented fully connected layers. The length, width, and height of the cubes and the bars in the Figure were drawn according to the dimensional size of data in each layer for a more intuitive display.