Learning and Using Taxonomies For Fast Visual Categorization

Abstract

The computational complexity of current visual categorization algorithms scales linearly at best with the number of categories. The goal of classifying simultaneously N_(cat) = 10^4 - 10^5 visual categories requires sub-linear classification costs. We explore algorithms for automatically building classification trees which have, in principle, log N_(cat) complexity. We find that a greedy algorithm that recursively splits the set of categories into the two minimally confused subsets achieves 5-20 fold speedups at a small cost in classification performance. Our approach is independent of the specific classification algorithm used. A welcome by-product of our algorithm is a very reasonable taxonomy of the Caltech-256 dataset.