Title: Semantic Hyperlapse: a recorder-aware and multi-importance approach for first person videos

The emergence of low-cost, high-quality personal wearable cameras combined with the increasing storage capacity of video-sharing websites have evoked a growing interest in first-person videos. Such videos are composed of long-running unedited streams which are usually tedious and unpleasant to watch; consequently, rise the need to provide quick access to the information therein. The State-of-the-art Semantic Hyperlapse method negligees the importance of the relevant information by only considering its significance, instead of analyzing how important it is. In this dissertation proposal, we propose a methodology to adaptively fast-forward First-Person Videos that emphasize the semantic portions applying a multi-importance approach. Further, we propose to address the dilemma of defining what the semantic information of a video is by considering its relevance to the recorder. Additionally, we analyze a novel Sparse Coding based adaptive frame selection methodology, and a smoothing process to address the discontinuity in the speed-up function. Preliminary results of the Multi-Importance approach lead to an improvement of 7 p.p. in the kept semantic content in the final video over the state-of-art method in Semantic Hyperlapse. We also achieved encouraging results for the other points of this proposal.

Prof. Erickson Rangel do Nascimento – Advisor (DCC – UFMG)
Prof. Mario Fernando Montenegro Campos – Co-advisor (DCC – UFMG)
Prof. André Vital Saúde (DCC – UFLA)
Prof. Arnaldo de Albuquerque Araújo (DCC – UFMG)
Prof. Mylene Christine Queiroz de Farias (ENE – UnB)
Prof. Ricardo da Silva Torres (IC – UNICAMP)