Skip to main content
. 2022 Apr 15;22(8):3049. doi: 10.3390/s22083049

Figure 5.

Figure 5

The overall architecture of the proposed model. In the architecture, we first use deep CNNs to extract the image vector features. Then, we construct graphs by calculating the distances between image vector features. Next is multiple principal neighborhood aggregation GCN. Each PNA-GCN is followed by a GRU layer. Finally, a MLP is used to classify features and images are labeled.