Postdoctoral position - CNAM Paris, France : Definition of a typology of multimedia data and evaluation of associated index structures
Keywords: multimedia data, data analysis, data indexing, content-based image and audio descriptors, multidimensional index structures
Duration: 1 year starting before end of 2008
Application deadline: September 30, 2008
Research groups and labs:
This post-doc will be conducted in the Vertigo research group of the French Laboratory in Computer Science CEDRIC at CNAM (Conservatoire National des Arts et Métiers) in Paris, in collaboration with the Database research group of LAMSADE Laboratory of Paris-Dauphine University and the Analysis-Synthesis research group of IRCAM Institute. CNAM: http://www.cnam.eu/ CNAM/CEDRIC/Vertigo: http://cedric.cnam.fr/AfficheEquipe.php?id=9&lang=en LAMSADE: http://www.lamsade.dauphine.fr/ (French) LAMSADE/Database research group: http://www.lamsade.dauphine.fr/groups.php?id_group=4 (French) IRCAM: http://www.ircam.fr/?L=1 IRCAM/Analysis-Synthesis research group: http://www.ircam.fr/anasyn.html?L=1
This work is supported by the DISCO project (2008-2010) which is a French initiative that gathers several research institutes and universities and that aims at designing and experimenting generic and flexible techniques for content-based indexing and searching, dedicated to distributed sources of multimedia documents (http://www.lamsade.dauphine.fr/rigaux/disco_anr/index).
Description:
The work consists in studying and defining a typology of multimedia data, and in evaluating several multidimensional index structures according to this typology. More precisely, it can be decomposed into two tasks: o The multimedia content-based descriptors (image, video and audio), made available by the consortium of the DISCO project, produce multidimensional features (signatures) of different natures. The associated spaces mainly differ in terms of dimensionality, size and distribution of the population. For example, content-based visual descriptors computed from frames of video sequences contain much more redundancy than those belonging to still images. The first objective of this work is to study the characteristics of such multidimensional spaces by proposing criteria that will allow defining a typology of these spaces. This study will facilitate the use and development of future index structures for rapid access to data, the underlying objective being to mutualize the work for several modalities. This part will be done in collaboration with Valerie Gouet-Brunet for image and video and with Geoffroy Peeters for audio contents. o The second task deals with the evaluation of index structures for large collections of signatures. The aim is to propose a framework for the evaluation of state-of-the-art index structures according to the proposed typology. For instance, previous works have demonstrated that several index structures are efficient with uniform distribution data (e.g. indexes based on space partitioning), whereas others are better with clustered distributions (e.g. tree-based indexes). Such a study will also conduct to the definition of criteria allowing the dynamic selection of the most appropriate indexing technique according to a given descriptor, to a given query type as well as to the potential combination of several modalities available to build the query. This part will be done in collaboration with Maude Manouvrier, Marta Rukoz and Valerie Gouet-Brunet.
Required skills:
PhD in computer science, databases and/or data analysis C/C++ or Java programming Experience in the problem of scalability (curse of dimensionality) or in non-textual data (image, video, sound)
Contacts:
Valerie Gouet-Brunet CNAM - CC 432, 292, rue Saint-Martin - F75141 Paris Cedex 03 Tel : +33 1 58 80 86 35/ Fax : +33 1 58 80 84 93 Valerie.Gouet@cnam.fr
Maude Manouvrier LAMSADE - Universite Paris IX Dauphine, Place du Maréchal De Lattre de Tassigny 75775 PARIS CEDEX 16 Tel : +33 1 44 05 41 85 / Fax : +33 1 44 05 40 90 manouvrier@lamsade.dauphine.fr
Geoffroy Peeters IRCAM - 1, pl. Igor Stravinsky 75004 Paris Tel : +33 1 44 78 14 22 / Fax : +33 1 44 78 15 40 Geoffroy.Peeters@ircam.fr
Marta Rukoz LAMSADE - Universite Paris IX Dauphine, Place du Maréchal De Lattre de Tassigny 75775 PARIS CEDEX 16 Tel : +33 1 44 05 41 85 / Fax : +33 1 44 05 40 90 Marta.Rukoz@dauphine.fr
Application procedure:
Before September 30, 2008, send your application by email to the four contacts: a detailed CV, a letter explaining your motivations for the topic and recommendation letters from three referees.