Multimodal Summarization of User-Generated Videos

The exponential growth of user-generated content has increased the need for efficient video summarization schemes. However, most approaches underestimate power aural features, while they are designed to work mainly on commercial/professional videos. In this work, we present an approach that uses both and visual features in order create summaries from Our produces dynamic summaries, is, comprising “important” parts original video, which arranged so as preserve their temporal order. We use supervised knowledge aforementioned modalities train a binary classifier, learns recognize important Moreover, novel dataset contains videos several categories. Every 1 s part each our been annotated by more than three annotators being or not. evaluate using classification strategies based audio, fused features. experimental results illustrate potential approach.