Abstract

The number of categories for action recognition is growing rapidly. It is thus becoming increasingly hard to collect suf-ficient training data to learn conventional models for each category. This issue may be ameliorated by the increasingly popular “zero-shot learning ” (ZSL) paradigm. In this frame-work a mapping is constructed between visual features and a human interpretable semantic description of each category, allowing categories to be recognised in the absence of any training data. Existing ZSL studies focus primarily on image data, and attribute-based semantic representations. In this pa-per, we address zero-shot recognition in contemporary video action recognition tasks, using semantic word vector space as the common space to embed videos and category labels. This is more challenging because the mapping between the semantic space and space-time features of videos containing complex actions is more complex and harder to learn. We demonstrate that a simple self-training and data augmenta-tion strategy can significantly improve the efficacy of this mapping. Experiments on human action datasets includ-ing HMDB51 and UCF101 demonstrate that our approach achieves the state-of-the-art zero-shot action recognition per-formance. Index Terms — action recognition, zero-shot learning 1.