On Tuesday October 30 Soledad Martin Fernandez presented her Master Final Work to obtain the title in the Master of Engineering in Telecommunication (@ masteresTicEpcc). The title: «University Life CBVR benchmark for evaluation system» under the direction of Manuel Barrena and Pablo García Rodríguez (GIM research group -Media Engineering Group- gim.unex.es). The tribunal were composed by Marisa Duran, Pedro Nuñez and Andres Caro.
This Master Final Work presents the development of a database of videos, designed to ease the process of evaluation of systems and video retrieval techniques based on content (CBVR: Content Based Video Retrieval).
CBVR systems use techniques for feature extraction by to compare videos and give similar results. These techniques extracted first, colors, shapes, movements … and on the other hand, objects or actions, which are displayed in the videos.
With this database we is intended to facilitate and improve the evaluation process of technic CBVR systems. The idea is to define the context in which it develops the content of the videos in the collection in order to carry out a careful and controlled entry of such content. A database cataloged in this way can be used as a battery of tests in experiments leading to the recovery of video-based content, such as cataloging to measure the quality of the response provided by the system. For example if we search our database videos relating to a library, the response can be assessed effectively simply by examining whether an element of the response is not as cataloged or «library» and a few of the items listed as «library «have been selected as relevant to the query.
To fulfill this purpose these databases should be tailored according to minimum standards of quality and properly labeled and cataloged. Throughout this work is explained the comprehensive study on the subject of videos, image recording and processing of all those terms used to describe audiovisual content in achieving this objective pursued.
Dialog box to enter the taxonomy in vManager (with MPEG-7):
CONTEXT AND OBJECTIVES:
The first chapter introduces the reader to the evolutionary line of research in content retrieval.
This field appeared due to the needs of search based on text documents, but the technology that surrounds us each time progresses more and gives us new devices such as camcorders, cameras and mobile, this are generating search needs according to new documents created image and video format. New approaches are needed to cover these modern concerns.
We present CBIR and CBIR systems for image retrieval based on its content and video retrieval based on content, respectively, focusing on the latter. We need creating video databases, aiming at the evaluation of the video retrieval system based on its content, CBVR.
It concludes with a set of goals to reach the desired goal with this work together with the contributions that we want us to share with this database of video.
Searching for Videos by keywords in the vManager system:
Graphical user interface to catalog videos database: