Abstract

Abstract — Current computational approaches to learning visual object categories require thousands of training images, are slow, cannot learn in an incremental manner and cannot incorporate prior information into the learning process. In addition, no algorithm presented in the literature has been tested on more than a handful of object categories. We present an method for learning object categories from just a few training images. It is quick and it uses prior information in a principled way. We test it on a dataset composed of images of objects belonging to 101 widely varied categories. Our proposed method is based on making use of prior information, assembled from (unrelated) object categories which were previously learnt. A generative probabilistic model is used, which represents the shape and appearance of a constellation of features belonging to the object. The parameters of the model are learnt incrementally in a Bayesian manner. Our incremental algorithm is compared experimentally to an earlier batch Bayesian algorithm, as well as to one based on maximum-likelihood. The incremental and batch versions have comparable classification performance on small training sets, but incremental learning is significantly faster, making real-time learning feasible. Both Bayesian methods outperform maximum likelihood on small training sets. I.

Citations

1371	Rapid object detection using a boosted cascade of simple features - Viola, Jones - 2001
1032	Object recognition from local scale-invariant features,” in Computer Vision - Lowe - 1999
656	An Introduction to Variational Methods for Graphical Models - Jordan, Ghahramani, et al. - 1999
646	Object class recognition by unsupervised scale-invariant learning - Fergus, Perona, et al.
612	A view of the EM algorithm that justifies incremental, sparse, and other variants - Neal, Hinton - 1998
550	Recognition-by-components: A theory of human image understanding - Biederman - 1987
222	Unsupervised learning of models for recognition - Weber, Welling, et al. - 2000
111	P.: A probabilistic approach to object recognition using local photometry and global geometry. In: ECCV - Burl, Weber, et al. - 1998
110	Learning parameters and structure of latent variables by variational Bayes - Attias - 1999
95	Local greyvalue invariants for image retrieval - Schmid, Mohr - 1997
77	A computational model for visual selection - Amit, Geman - 1999
75	A Statistical Approach to 3D Object Detection Applied to Faces and Cars (cmu-ri-tr-00-06 - Schneiderman - 2000
73	Learning from one example through shared densities on transforms - Miller, Matsakis, et al. - 2000
56	Bayesian Methods for Mixtures of Experts - Waterhouse, MacKay, et al. - 1996
47	saliency and image description - Kadir, Brady - 2001
3	A bayesian approach to unsupervised learning of object categories - Fei-Fei, Fergus, et al. - 2003