Retrieval knowledge graph library generation method based on combination of scene graph and concept network

A retrieval knowledge graph library generation method based on combination of a scene graph and a conceptual network comprises the following steps: 1) model pre-training: pre-training input data on a neural network, and detecting types and positions of objects appearing in a picture; 2) training a scene graph: carrying out unbiased training on a pre-training result of the model, finally outputting a file of information related to the scene graph by applying a neural network model, and predicting a relationship between different types in the image; 3) automatically expanding the knowledge graph; 4) testing the trained scene graph model; 5) extracting and processing file information related to the scene graph and the concept network, and then importing the file information into a retrieval database to finally form the retrieval database; combining nodes and relations with high similarity in the scene graph, fusing the scene graph and the knowledge graph database corresponding to the concept network, and finally