Multiple instance boosting for face recognition in videos

Abstract

For face recognition from video streams often cues such as transcripts, subtitles or on-screen text are available. This information could be very valuable for improving the recognition performance. However, frequently this data can not be associated directly with just one of the visible faces. To overcome this limitations and to exploit valuable information, we define the task as a multiple instance learning (MIL) problem. We formulate a robust loss function that describes our problem and incorporates ambiguous and unreliable information sources and optimize it using Gradient Boosting. A new definition of the posterior probability of a bag, based on the L p -norm, improves the ability to deal with varying bag sizes over existing formulations. The benefits of the approach are demonstrated for face recognition in videos on a publicly available benchmark dataset. In fact, we show that exploring new information sources can drastically improve the classification results. Additionally, we show its competitive performance on standard machine learning datasets. © 2011 Springer-Verlag.