Knowledge-Based Scene Graph Generation in Medical Field

This work aims to combine object detection and knowledge graphs to understand medical scenes. The rapid growth of the medical field has introduced various equipment types which makes scene analysis more crucial. By representing object detection results as a knowledge graph, one can derive conclusions and gain a clear understanding of the scene through scene graphs. While object detection identifies objects in images, it lacks the ability to utilize contextual information and visual reason. Therefore, knowledge-aware object detection becomes essential in allowing the integration of external knowledge graphs into object detection algorithms. Therefore, the work focuses on the medical field that combines object detection and object relationships to form a correlated understanding of the scene. Construct relationships between objects, generate scene graphs, and derive inferences to predict the overall scene understanding is achieved through the knowledge graph. The proposed work incorporates knowledge graphs in the medical domain that addresses the need for improved scene understanding in medical contexts.