SAVASA project@ TRECVid 2013: semantic indexing and interactive surveillance event detection
Suzanne Little, Iveel Jargalsaikhan, Rami Albatal, Cem Direkoğlu, Noel Edward O'Connor, Alan F. Smeaton, Kathy Clawson, Min Jing, Brian Scotney, Hui Wang · Dublin City University Open Access Institutional Repository (Dublin City University) · 2013
In this paper we describe our participation in the semantic indexing (SIN) and interactive surveillance event detection (SED) tasks at TRECVid 2013 [11]. Our work was motivated by the goals of the EU SAVASA project (Standards-based Approach to Video Archive Search and Analysis) which supports search over multiple video archives. Our aims were: to assess a standard object detection methodology (SIN); evaluate contrasting runs in automatic event detection (SED) and deploy a distributed, cloud-based search interface for the interactive component of the SED task. Results from the SIN task, underlying retrospective classifiers for the surveillance event detection and a discussion of the contrasting aims of the SAVASA user interface compared with the TRECVid task requirements are presented.