Learning Large Scale Class Specific Hyper Graphs for Object Recognition

This paper describes how to construct a hyper-graph model from a large corpus of multi-view images using local invariant features. We commence by representing each image with a graph, which is constructed from a group of selected SIFT features. We then propose a new pairwise clustering method based on a graph matching similarity measure. The positive example graphs of a specific class accompanied with a set of negative example graphs are clustered into one or more clusters, which minimize an entropy function with a restriction defined on the F-measure. Each cluster is simplified into a tree structure composed of a series of irreducible graphs, and for each of which a node cooccurrence probability matrix is obtained. Finally, a recognition oriented class specific hyper-graph (CSHG) is generated from the given graph set. Experiments are performed on over 50K training images spanning 500 objects and over 20K test images of 68 objects. This demonstrates the scalability and recognition performance of our model.