Naive bayes image classification: beyond nearest neighbors

Abstract. Naive Bayes Nearest Neighbor (NBNN) has been proposed as a powerful, learning-free, non-parametric approach for object classifi-cation. Its good performance is mainly due to the avoidance of a vector quantization step, and the use of image-to-class comparisons, yielding good generalization. In this paper we study the replacement of the near-est neighbor part with more elaborate and robust (sparse) representa-tions, as well as trading performance for speed for practical purposes. The representations investigated are k-Nearest Neighbors (kNN), Iter-ative Nearest Neighbors (INN) solving a constrained least squares (LS) problem, Local Linear Embedding (LLE), a Sparse Representation ob-tained by l1-regularized LS (SRl1), and a Collaborative Representation obtained as the solution of a l2-regularized LS problem (CRl2). In partic-ular, NIMBLE and K-DES descriptors proved viable alternatives to SIFT and, the NBSRl1 and NBINN classifiers provide significant improve-ments over NBNN, obtaining competitive results on Scene-15, Caltech-101, and PASCAL VOC 2007 datasets, while remaining learning-free approaches (i.e., no parameters need to be learned).