Research on Visual Relation Detection Based on Computer Vision

摘要

Extracting semantic information from unstructured data, such as images or text, is a critical challenge in artificial intelligence and a long-lasting research direction. In general, semantic information is captured in more meaningful ways by the relationship between objects. In particular, visual relationship can be represented by triples of this form (subject, predicate, object). Visual relationship detection needs to identify objects in the image and their location, as well as identify relationships between objects. In this paper, we review the methods about visual relation detection, analyses the algorithms at home and abroad, and compare the advanced and inadequacies between algorithms. Finally, we look forward to the development trend of visual relation detection.