Boosting Separability in Semisupervised Learning for Object Classification

Boosting algorithms, especially AdaBoost, have attracted great attention in computer vision. In the early version of boosting algorithms, the weak classifier selection and the strong classifier learning are linked together. It has been demonstrated that decoupling of these two processes can provide more flexibility for training a better classifier. In these studies, linear discriminant analysis (LDA) has been adopted to select weak classifiers independently based on class separability rather than a training error that occurs normally in AdaBoost. It is observed that LDA is successful only if a large number of labeled training samples is available. However, a large-scale labeled training set is not always available in many computer vision applications such as object classification. To tackle this problem, this paper proposes semisupervised subspace learning combined with a boosting framework for object classification, through which unlabeled data can participate in the boosting training to compensate for the lack of enough labeled data. With the proposed framework, this paper develops three various approaches that utilize unlabeled data in different ways. According to the experiments on several public image data sets, the proposed methods achieve superior performance over AdaBoost and existing semisupervised algorithms.

I. Introduction

As one of the widely used ensemble-based learning algorithms, boosting has achieved great successes on various computer vision applications (e.g., object detection [1], object classification [2], object tracking [3], image retrieval [4]) since the first real-time face detector [5]. The strategy of boosting is to progressively combine weak classifiers to form a strong classifier, instead of trying to learn a single complex classifier. During the training process, boosting pays much more attention to those misclassified training samples [6]. As a general and effective learning scheme, boosting has been deeply studied in theory and has been adopted to deal with a wide variety of problems in pattern recognition. Briefly, two explanations are proposed to understand the success of boosting: functional gradient view [7] and margin-based optimization [8]. Given the loss function of boosting, the former finds the best weak classifier by gradient descent in each step, and its combination weight can be obtained simultaneously. The latter tries to formulate this problem as a margin-based optimization problem as support vector machine (SVM) does. A complete review of boosting algorithms is presented in [9].