Utilizing Co-occurrence Patterns for Semantic Concept Detection in Images
Linan Feng, Bir Bhanu · 2015
Semantic concept detection is an important open problem in concept-based image understanding. In this paper, we develop a method inspired by social network analysis to solve the semantic concept detection prob-lem. The novel idea proposed is the detection and utilization of concept co-occurrence patterns as con-textual clues for improving individual concept detec-tion. We detect the patterns as hierarchical communi-ties by graph modularity optimization in a network with nodes and edges representing individual concepts and co-occurrence relationships. We evaluate the effect of detected co-occurrence patterns in the application sce-nario of automatic image annotation. Experimental re-sults on SUN’09 and OSR datasets demonstrate our ap-proach achieves significant improvements over popular baselines. 1.