Joint SVM for Accurate and Fast Image Tagging

This paper studies how joint training of multiple support vector machines (SVMs) can improve the effectiveness and efficiency of automatic image annotation. We cast image annotation as an output-related multi-task learning framework, with the prediction of each tag’s presence as one individual task. Evidently, these tasks are related via correlations between tags. The proposed joint learning framework, which we call joint SVM, can encode the correlation between tags by defining appropriate kernel functions on the outputs. Another practical merit of the joint SVM is that it shares the same computational complexity as one single conventional SVM, although multiple tasks are solved simultaneously. According to our empirical results on an image-annotation benchmark database, our joint training strategy of SVMs can yield substantial improvements, in terms of both accuracy and efficiency, over training them independently. In particular, it outperforms many other state-of-the-art algorithms.