Generalized Class Representative Computation with Graph Embedding and Clustering
Demirci, Muhammed Fatih
MetadataShow full item record
In this paper, we propose an object category representation framework by first showing objects as graph structures and embedding graphs into vector spaces for color object recognition. The object categories are then built by clustering. The distance between an object and an object category is computed by Earth Mover's Distance. The proposed method has been successfully evaluated on a number of datasets, and its performance has been computed against previous techniques.