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

Figure 1

Pooling effect diagram. The left side of the diagram is a 4 × 4 feature map that consists of 2 × 2 pooling windows with a step length of 2 for downsampling the feature map. The right side shows the results after mean pooling, namely, the average value of the selected area is used as the pooled result. For max pooling, the pooled result would be the maximum value of the selected area.