The ObjectCube Project: Multi-Dimensional Media Browsing

Since the introduction of personal computers, and digital recording devices in particular, personal collections of digital media have been growing ever larger. A typical personal computer now contains a multitude of files containing various documents, photos, videos, and music. Keeping track of the location and contents of all these files is turning into a major headache for most people and users find it increasingly difficult to organize and retrieve the contents of their computer. It is therefore increasingly important to provide effective tools for browsing personal media collections.

The MDA model was instrumental in helping users make sense of vast amounts of numerical data. The model is implemented through browsers that help users discover their data collections without bothering them with location. These browsers are of great practical interest because they focus on the value of data items as well as on the relationships that exist between data, abstracting physical considerations. Our goal is to apply the concepts of the MDA model to media browsing. This allows us to define in a very clear manner the concepts according to which media items can be grouped, the way tags attached to media files are organized, typically in well-defined hierarchies, or the way users switch from one set of documents to the other when navigating between concepts, hierarchies, or levels of detail. While the MDA model is not immediately suited to media browsing, we believe that the similarity of the problems is significant enough to use the concepts of that model as the foundation of a powerful and flexible solution to media browsing.

ObjectCube is a generic multi-dimensional analysis model for media files. The model is based on the concepts of multi-dimensional analysis (MDA) which have been successfully used in OLAP applications to view and analyze data. We have developed a prototype media server implementing the ObjectCube model, for the purpose of demonstrating the efficiency of the model as well as for the future development of browsing interfaces based on the model; the server will be made publically available.

We have also developed a prototype 3D image browser, based on ObjectCube, called PhotoCube. PhotoCube has three browsing dimensions, some or all of which can be used simultaneously. Using the ObjectCube data model, it can also pivot dimensions on and off the screen, resulting in practical multi-dimensional image browsing.

Screenshots  (Click on the thumbnails for full sized images.)

The following screenshots come from PhotoCube.


The following videos come from PhotoCube.
  1. Beautiful photos, information discovery using hiking photos.
  2. A data retrieval and information discovery scenario, using family photos.
  3. A story telling scenario, using hiking photos.
  4. A data filtering scenario, using hiking photos.
  5. An information discovery scenario, drilling down, rolling back and pivoting; using animal photos.
  6. A pattern discovery scenario, using hiking photos and focused on volcanic ash from Eyjafjallajökull.

Publications (updated February 2012)

Grímur Tómasson, Björn Þór Jónsson, Laurent Amsaleg.
ObjectCube: A Novel Data Model for Multi-Dimensional Media Browsing.
Technical Report RUTR-CS12002, Reykjaví­k University, January, 2012.
Submitted for conference publication.

Hlynur Sigurþórsson, Grímur Tómasson, Björn Þór Jónsson, Laurent Amsaleg.
PhotoCube: Towards Multi-Dimensional Image Browsing.
Technical Report RUTR-CS12001, Reykjaví­k University, January, 2012.
Submitted for conference publication.

Kristján Rúnarsson.
A Face Recognition Plug-in for the PhotoCube Browser.
M.Sc. Thesis. Reykjaví­k University. December, 2011.

Hlynur Sigurþórsson.
Photocube: A Multi-Dimensional Image Browser.
M.Sc. Thesis. Reykjaví­k University. June, 2011.

Grímur Tómasson, Hlynur Sigurþórsson, Björn Þór Jónsson, Laurent Amsaleg.
PhotoCube: Effective and Efficient Multi-Dimensional Browsing of Personal Photo Collections (demonstration paper).
In ACM International Conference on Multimedia Retrieval (ICMR), Trento, Italy, April, 2011.

Grímur Tómasson.
ObjectCube - A Generic Multi-Dimensional Model for Media Browsing.
M.Sc. Thesis. Reykjaví­k University. January, 2011.