Knowledge Graph Representation Fusion Framework for Fine-Grained Object Recognition in Smart Cities

Figure 2

Overview of our KGFR framework. It consists of a visual feature extraction module, a knowledge expression module, and a knowledge fusion module. In visual feature extraction module, the CNN can extract the overall visual features in images. In the knowledge representation module, the constructed knowledge graph is split into two subgraphs, and the representations of the nodes are obtained by two GATs, respectively. Finally the knowledge representations of the same nodes in different GATs are concatenated to obtain the final knowledge representation. In the knowledge fusion module, the MCB module is introduced to fuse the knowledge representations with the extracted visual features, so as to enhance the fine-grained image classification.