Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence (AAAI-17) Event Video Mashup: From Hundreds of Videos to Minutes of Skeleton Lianli Gao,1 Peng Wang,2 Jingkuan Song,3 Zi Huang,2 Jie Shao,1 Heng Tao Shen1 1University of Electronic Science and Technology of China, Chengdu 611731, China. 2The University of Queensland, QLD 4072, Australia. 3Columbia University, NY 10027, USA. {lianli.gao, shaojie}@uestc.edu.cn, {p.wang6, huang}@itee.uq.edu.au,
[email protected],
[email protected] Abstract 1J1. 1JV `V QJ`V`VJHV %]VH VG 1RVQ VC: VR Q 8Q QJ :`: .QJ QIG1J$9 The explosive growth of video content on the Web has been revolutionizing the way people share, exchange and perceive :.%] information, such as events. While an individual video usu- 1IV 1JV ally concerns a specific aspect of an event, the videos that are uploaded by different users at different locations and times can embody different emphasis and compensate each other :%:C 1V G:I: 1H 1I in describing the event. Combining these videos from dif- ferent sources together can unveil a more complete picture of the event. Simply concatenating videos together is an in- Figure 1: Overview of Event Video Mashup tuitive solution, but it may degrade user experience since it is time-consuming and tedious to view those highly redun- dant, noisy and disorganized content. Therefore, we develop ods usually rely on low-level cues to determine the impor- a novel approach, termed event video mashup (EVM), to au- tomatically generate a unified short video from a collection of tance of segments of a video (Ma et al.