Overall Objectives
Scientific Foundations
Application Domains
New Results
Contracts and Grants with Industry
Other Grants and Activities

Section: Software


Automatic Generation of Hypervideos

Participants : Sébastien Campion [ correpondent ] , Mathieu Ben, Camille Guinaudeau, Gwénolé Lecorvé.

This work was made with the help of Guillame Gravier, from the Metiss project-team.

We created a demonstrator to illustrate an application of video topic segmentation on a collection of TV news programs (INA corpus) in collaboration with Guillaume Gravier from the Metiss project-team. The core of the system is our topic segmenter which is fed by the output of an automatic speech transcriptor, and the output of the S2E module dedicated to automatic extraction of structuring events in the video. Behind the topic segmenter, the Kiwi module extracts a list of keywords from each topic segment. Using these keywords we then create links to web pages dealing with the same topic, and links to related video segments inside the collection. All the generated metadata for a given video are used to generate a web page, that we call a hypervideo, and which allows non-linear browsing of the video, according to topic segments. Furthermore, the user can jump to web pages related to his/her topic of interest, or to other reports in the video collection dealing with the same or similar topics. Each time, the user can play the corresponding video segment in a player fully integrated in the browser. To this aim, we used the last version (3.5) of the Firefox browser which handles video HTML mark-ups.

This demo was presented at the NEM summit 2009, St-Malo, France.

Image Search Engines Comparator

Participants : Sébastien Campion [ correpondent ] , Laurent Amsaleg, Nguyen Khang Pham.

Using ISEC (Image Search Engine Comparator), we publish a website which gives the possibility to use and compare several CBIR search engines. Currently we can use NVTree search engine on several datasets (up to 10 millions of images) and IRCA (Image Retrival by Correspondence Analysis) search engine.


Logo Inria