LAUSR

In this paper, we focus on detecting highlights in online videos. Given the explosive growth of online videos, it is becoming increasingly important to single out those highlights for audiences instead of requiring them browsing every tedious part of the video. It is ideally that the contents of extracted highlights can be consistent with the topic of the video as well as the preference of the individual audience. To this end, this paper introduces a novel content-aware approach by formulating the highlights detection in a transfer learning framework. Under this framework. The experimental results on three different types of videos show that our content-aware highlight extraction method is particularly useful for online videos content fetching, e.g. showing the abstraction of the entire video while playing focus on the parts that matches the user queries.