Snapshots from MultiMediaMiner


MultiMediaMiner data warehouse with cubes and dimensions:
The figure shows the user interface of the MultiMediaMiner data warehouse with 4 data cubes and the dimensions and measurements defined in one of them. ColorCube contains dimentions pertaining to colour and texture, SizeCube contains mainly numeric dimensions like image size, width, height, etc., while PopCube contains dimensions related to popularity of images and richness of web pages. BigCube contains most of the dimensions together.


Selecting (and browsing) data sets of images using keyword hierarchy:
Thumbnails of images and video frames in the database can be browsed with MultiMediaMiner user interface. In this example thumbnails of commercial airplanes pertaining to the airliner Boeing are displayed. This user interface also allows the selection of a multimedia data set to be mined. The hierarchy of keywords on the left is a section of the concept hierarchy automatically generated by visiting some web sites containing aircraft images. By selecting a keyword, the images associated with the word and its descendents are selected. By clicking on a thumbnail, the image from the web and a list of links to the web pages that contain it are displayed.


Browsing 3 dimensions of the multimedia data cube:
A visualization tool used to browse multi-dimensional data cubes, 3 dimensions at a time. The concept hierarchy defined on each dimension allows drilling-down and rolling-up along a dimension. This type of data cube browsing gives a big picture of the content of the database and even allows to see rough clustering of data values. Selecting a sub-cube from the view drills through it up to the raw data, and one can see the set of multimedia items in the selected sub-cube and even the web pages that contain them. 


Visualization of association rules:
A visualization of some association rules. The existence of a column on the grid represents an association between the left-hand side parameters and the right-hand side parameters. The height of the column depicts the support of the association rule it represents while the colour of the column describes the confidence of the rule. 


MultiMediaMiner Characterizer:
The view describes in a histogram graph the general characteristics for two dimensions: the size of the media in bytes and the Internet domain from which the media was extracted. For this example, only three Internet domains were considered, while the sizes were ``rolled-up'' to a higher concept of media size (i.e. small, medium and large). With this user interface, it is possible to visualize any two dimensions at a time, and drill-down or roll-up along a given dimension to find characteristics on more concrete values or specialized concepts. 


MultiMediaMiner Classifier user interface:
The figure shows an output of the classifier module where a classification of images and frames based on their width, height, and file size, with reference to the distribution of image format, is made for a given Web site. By clicking on a class, it is possible to drill through to the raw data.