My research interests include inter-modal analysis of multimedia documents in large databases, image and video processing, indexing and retrieval, data mining, knowledge representation, and information retrieval models.

Inter-modal analysis consists in modeling and learning some relationships between several modalities from a data set in order to extract semantics. In the task of document annotation - namely in my case of visual and textual modalities, extracted semantics is used to predict a textual annotation given a visual information. I am more specifically interested in the automatic annotation of image and video documents. I investigate several approaches of annotation of large news video databases, including TRECVID collection and our local Japanese news archives.

In my previous research (PhD works), I defined an image retrieval model suited to visual documents, integrating a rich semantics for representing both visual data and users' queries, while allowing the application to large databases. This model merges the vector space model of information retrieval - a widely tested model in text retrieval - with the conceptual graph formalism. The model is based on the use of star graphs (elementary conceptual graphs made up of a single relation connected to some concepts representing image objects) as image descriptors. Experiments showed that integrating relations into the vector space model with star graphs increases the system precision, and that the results are comparable to graph projection systems, while decreasing the processing time for users' queries.

Another contribution of my PhD works is the definition and the validation of a novel weighting scheme for image objects. This weighting scheme has been designed according to human perception criteria about what is estimated as important in images. Four basic hypotheses related to human perception have been defined and their validity has been estimated as compared to actual observations from a user study. This lead to a formal definition of the weighting scheme, which has been proved consistent with the user study.

Since Dec. 2005: JSPS Postdoctoral Fellow (24 months), Multimedia Information Research division (National Institute of Informatics) Tokyo, Japan.
Oct. 2004-Sept. 2005: Research-teaching assistant (12 months), CLIPS-IMAG, Grenoble, France.
Oct. 2001-Dec. 2004: PhD candidate (3 years), CLIPS-IMAG, Grenoble, France.
Jul. 2002-Dec. 2002: Visiting researcher (6 months), Information Retrieval Group, Glasgow, UK.
Mar. 2001-Sept. 2001: Internship (6 months), IPAL, National University of Singapore, Singapore:

JSPS Postdoctoral fellowship for foreign researcher, December 2005-November 2007.

Eurodoc program grant (Region Rhone-Alpes), July 2002-December 2002.

2007

(submitted) Multimedia objects for user-centered similarity matching, Jean Martinet, Yves Chiaramella, Philippe Mulhem, Shin'ichi Satoh, submitted to Multimedia Tools and Applications Journal, Special Issue on ``Semantic Multimedia''. (view pdf)

Image-based Quizzes from News Video Archives, Masanori Sano, Nobuyuki Yagi, Jean Martinet, Norio Katayama and Shin'ichi Satoh, IEEE International Conference on Multimedia & Expo (ICME'07), Beijing International Convention Center, Beijing, China, pp-, July 2-5, 2007 (view pdf)

A study of intra-modal association rules for visual modality representation, Jean Martinet, Shin'ichi Satoh, submitted to the fifth IEEE International Workshop on Content-Based Multimedia Indexing (CBMI'07), Bordeau, France, pp -, June 25-27, 2007 (view pdf)

Using visual-textual mutual information for inter-modal document indexing, Jean Martinet, Shin'ichi Satoh, 29th BCS-IRSG European Conference on Information Retrieval Research (ECIR'07), Rome, Italy, pp -, April 2-5 2007 (view pdf)

An information theoretic approach for automatic document annotation from inter-modal analysis, Jean Martinet, Shin'ichi Satoh, 20th International Joint Conference on Artificial Intelligence (IJCAI-2007), Workshop on Multimodal Information Retrieval, Hyderabad, India - January 6, 2007 (view pdf )

2006

Annotation automatique de documents par analyse intermodale, Jean Martinet, Journées Scientifiques Francophones (JSF'06), Tokyo, Japan - December 1, 2006 (in French)

Statistical correlation of visual data and semantics, a brief survey, Jean Martinet, Internal report, National Institute of Informatics, March 29, 2006

2005

A Model for Weighting Image Objects in Home Photographs, Jean Martinet, Yves Chiaramella, Philippe Mulhem, 14th International Conference on Information and Knowledge Management (CIKM'05), Bremen, Germany, pp760-767, 31 October - 5 November, 2005 (view pdf)

A relational vector-space model of information retrieval adapted to images, Jean Martinet, SIGIR Forum 39(1):62, June 2005

2004

Un modèle vectoriel relationnel de recherche d'information adapté aux images, Jean Martinet, Ph.D. thesis, Université Joseph Fourier, Grenoble, France, December, 2004 (in French) (view pdf)

2003

Photograph indexing and retrieval using star-graphs, Jean Martinet and Yves Chiaramella and Philippe Mulhem and Iadh Ounis, Content-Based Multimedia Indexing (CBMI'03), Rennes, France, pp335-341, 22-24 September, 2003 (view pdf)

A weighting scheme for star-graphs, Jean Martinet and Iadh Ounis and Yves Chiaramella and Philippe Mulhem, 25th BCS-IRSG European Conference on Information Retrieval Research (ECIR'03), Pisa, Italy, pp 546-554, April, 2003 (view pdf)

2002

Un modèle vectoriel étendu de recherche d'information adapté aux images, Jean Martinet and Yves Chiaramella and Philippe Mulhem, 20ème Congrès INFORSID'02 (Informatique des Organisations et Systèmes d'Information et de Décision), Nantes, France, pp337-348, 4-7 June, 2002 (in French) (view pdf)

2001

Un modèle vectoriel de recherche d'information pour les images, Jean Martinet, M.Sc. thesis, MRIM - CLIPS-IMAG, Grenoble, France, June, 2001 (in French).