Learning and using taxonomies for fast visual categorization

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, logNcat 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.