Viewpoint-aware Video Summarization

This paper introduces a novel variant of video summarization, namely building a summary that depends on the particular aspect of a video the viewer focuses on. We refer to this as $\mathit{\text{viewpoint}}$. To infer what the desired $\mathit{\text{viewpoint}}$ may be, we assume that several other videos are available, especially groups of videos, e.g., as folders on a person’s phone or laptop. The semantic similarity between videos in a group vs. the dissimilarity between groups is used to produce $\mathit{\text{viewpoint}}$-specific summaries. For considering similarity as well as avoiding redundancy, output summary should be (A) diverse, (B) representative of videos in the same group, and © discriminative against videos in the different groups. To satisfy these requirements (A)-© simultaneously, we proposed a novel video summarization method from multiple groups of videos. Inspired by Fisher’s discriminant criteria, it selects summary by optimizing the combination of three terms (a) inner-summary, (b) inner-group, and © between-group variances defined on the feature representation of summary, which can simply represent (A)-©. Moreover, we developed a novel dataset to investigate how well the generated summary reflects the underlying $\mathit{\text{viewpoint}}$. Quantitative and qualitative experiments conducted on the dataset demonstrate the effectiveness of proposed method.