High-level attributes modeling for indoor scenes classification

Scene classification is a challenging problem in computer vision. Though conventional methods show good performance in recognizing outdoor scenes, these methods does not work well in indoor scenes recognition. In recent years, high level image representations consisted of semantic attribute information has been introduced to solve this problem. However, a key technical challenge for these representations is the "curse of dimensionality", caused by the large numbers of objects and high dimensionality of the response vector for each object. In this paper, we propose a hypergraph learning algorithm based feature selection method for indoor scene classification. It performs feature selection by hypergraph regularization, which not only considers the interaction among features but also the interaction between the feature selection heuristics and the corresponding classifier. For the convenience of the prediction of the new images, a liner regression model is integrated in the framework, making the new images classification directly and in real time. The experimental results show that our approach has satisfactory performance compared with previously proposed methods.