Semantic Concept Co-occurrence Patterns for Image Annotation and Retrieval,Describing visual image contents by semantic concepts is an effective and straightforward way to facilitate various high level applications.Representing images by semantic concepts instead of visual features remains a challenging problem. Generating semantic descriptors manually is not feasible due to the ever-growing number of image collections.

Current machine intelligence and statistical learning techniques for inferring semantic concepts from low-level features struggle in bridging the semantic gap. However, many image-based applications such as retrieval, annotation, recommendation, indexing and ranking, require an effective semantic representation of images.

There is a growing need in automatically inferring concepts from visual properties by learning the correspondence from loosely labeled data.