TR2012-026

Scalable Active Learning for Multi-Class Image Classification


Abstract:

Machine learning techniques for computer vision applications like object recognition, scene classification, etc. require a large number of training samples for satisfactory performance. Especially when classification is to be performed over many categories, providing enough training samples for each category is infeasible. This paper describes new ideas in multi-class active learning to deal with the training bottleneck, making it easier to train large multi-class image classification systems. First we propose a new interaction modality for training which requires only yes-no type binary feedback instead of a precise category label. The modality is especially powerful in the presence of hundreds of categories. For the proposed modality, we develop a Value-of-Information (VOI) algorithm that chooses informative queries while also considering user annotation cost. Second, we propose an active selection measure that works with many categories and is extremely fast to compute. This measure is employed to perform a fast seed search before computing VOI, resulting in an algorithm that scales linearly with data-set size. Third, we use locality sensitive hashing to provide a very fast approximation to active learning, which gives sub-linear time scaling allowing application to very large data-sets. The approximation provides up to two orders of magnitude speedups with little loss in accuracy. Thorough empirical evaluation of classification accuracy, noise sensitivity, imbalanced data, and computational performance on a diverse set of image data-sets demonstrates the strengths of the proposed algorithms.

 

  • Related News & Events