... creation from human language scenario.
Séminaire de M. Sébastien Poullot, project researcher, NII, Tokyo, Japan - Date : mercredi 18 juillet 2012, 10h00 - Lieu : Seminar room, 9ème étage, B1, MICA Institute, Hanoi University of Science and Technology

Intervenant :
M. Sébastien POULLOT, Project researcher, National Institute of Informatics NII, Tokyo, Japan

Date : mercredi 18 juillet 2012, 10h00
Lieu : Institut MICA,; salle "seminar room", 9ème étage, bâtiment B1, Institut Polytechnique de Hanoi
Interprète traducteur : le séminaire sera présenté en anglais

Résumé/abstract:
Scalable Content Based Video Copy Detection
This talk addresses the video similarity search (here copies) at very large scale: video databases of 500,000 hours and more. It explores different aspects of the constraints so as to be able to monitor a stream with a single core in delayed real time. The stress of the talk is on the Z-grid indexing scheme and the probabilistic search.

Video mining (and extension to image and object search)
Based of video copy detection, we can perform some mining in large video database. But from a linear problem we come to a quadratic problem. In order to lower the complexity of the problem another another description level is proposed along with a new indexing scheme.

Automatic video creation from human language scenario
This talk is a proposition of a concrete application. It aims to create, from existing contents (web and personal collections), short movies which scenario is typed by a lambda user. It is a prospective talk which explain the processes chain and theirs locks. Participants are very welcome to express their opinion, and to share ideas.