Search

Existing action recognition algorithms require a set of positive exemplars to train a classifier for each action. However, the amount of action classes is very large and the users' queries vary dramatically. It is impractical to pre-define all possible action classes beforehand. To address this issue, we propose to perform action recognition with no positive exemplars, which is often known as the zero-shot learning. Current zero-shot learning paradigms usually train a series of attribute classifiers and then recognize the target actions based on the attribute representation. To ensure the maximum coverage of ad-hoc action classes, the attribute-based approaches require large numbers of reliable and accurate attribute classifiers, which are often unavailable in the real world. In this paper, we propose an approach that merely takes an action name as the input to recognize the action of interest without any pre-trained attribute classifiers and positive exemplars. Given an action name, we first build an analogy pool according to an external ontology, and each action in the analogy pool is related to the target action at different levels. The correlation information inferred from the external ontology may be noisy. We then propose an algorithm, namely adaptive multi-model rank-preserving mapping (AMRM), to train a classifier for action recognition, which is able to evaluate the relatedness of each video in the analogy pool adaptively. As multiple mapping models are employed, our algorithm has better capability to bridge the gap between visual features and the semantic information inferred from the ontology. Extensive experiments demonstrate that our method achieves the promising performance for action recognition only using action names, while no attributes and positive exemplars are available. [ABSTRACT FROM AUTHOR]

Abstract: Agglomerative clustering, which iteratively merges small clusters, is commonly used for clustering because it is conceptually simple and produces a hierarchy of clusters. In this paper, we propose a novel graph-structural agglomerative clustering algorithm, where the graph encodes local structures of data. The idea is to define a structural descriptor of clusters on the graph and to assume that two clusters have large affinity if their structural descriptors undergo substantial change when merging them into one cluster. A key insight of this paper to treat a cluster as a dynamical system and its samples as states. Based on that, Path Integral, which has been introduced in statistical mechanics and quantum mechanics, is utilized to measure the stability of a dynamical system. It is proposed as the structural descriptor, and the affinity between two clusters is defined as Incremental Path Integral, which can be computed in a closed-form exact solution, with linear time complexity with respect to the maximum size of clusters. A probabilistic justification of the algorithm based on absorbing random walk is provided. Experimental comparison on toy data and imagery data shows that it achieves considerable improvement over the state-of-the-art clustering algorithms. [Copyright &y& Elsevier]

In this paper, linear local tangent space alignment (LLTSA), as a novel linear dimensionality reduction algorithm, is proposed. It uses the tangent space in the neighborhood of a data point to represent the local geometry, and then aligns those local tangent spaces in the low-dimensional space which is linearly mapped from the raw high-dimensional space. Since images of faces often belong to a manifold of intrinsically low dimension, we develop LLTSA algorithm for effective face manifold learning and recognition. Comprehensive comparisons and extensive experiments show that LLTSA achieves much higher recognition rates than a few competing methods. [Copyright &y& Elsevier]