Detecting interesting events using unsupervised density ratio estimation

摘要: Generating meaningful digests of videos by extracting interesting frames remains a difficult task. In this paper, we define events as unusual which occur rarely in the entire video and propose novel event summarization framework based on technique density ratio estimation recently introduced machine learning. Our proposed is unsupervised it can be applied to general sources, including from moving cameras. We evaluated approach publicly available dataset context anomalous crowd behavior with challenging personal dataset. demonstrated competitive performance both accuracy relative human annotation computation time.