퐾-NN BOOSTING PROTOTYPE LEARNING FOR OBJECT CLASSIFICATION

Object classification is a challenging task in computer vision. Many approaches have been proposed to extract meaning-ful descriptors from images and classifying them in a super-vised learning framework. In this paper, we revisit the clas-sic 푘-nearest neighbors (푘-NN) classification rule, which has shown to be very effective when dealing with local image de-scriptors. However, 푘-NN still features some major draw-backs, mainly due to the uniform voting among the near-est prototypes in the feature space. In this paper, we pro-pose a generalization of the classic 푘-NN rule in a supervised learning (boosting) framework. Namely, we redefine the vot-ing rule as a strong classifier that linearly combines predic-tions from the 푘 closest prototypes. To induce this classifier, we propose a novel learning algorithm, MLNN (Multiclass Leveraged Nearest Neighbors), which gives a simple proce-dure for performing prototype selection very efficiently. We tested our method on 12 categories of objects, and observed significant improvement over classic 푘-NN in terms of classi-fication performances. 1