The generic object categorization using the Latent Dirichlet allocation model and bag of Biologically Inspired Model features

The generic image categorization was a challenging problem because of the wide variety of objects in image. In this paper, we proposed a method using the hybrid generative/discriminative approach combining the Biologically Inspired Model (BIM) to implement the generic object categorization. The main contributions were below: 1) We proposed an feature extraction method, which adjust BIM, and formed bag of BIM(BOBIM) feature. 2) We used LDA model to extract the semantic topics, and used SVM to make final decision. The LDA/SVM model was a hybrid generative/discriminative approach. The experimental results reveal the efficiency of our method.