Support page for MUBI
This applications requires the Java Run-time Environment (JRE) version 1.4. The latest version of the JRE can be found at Sun's Java web-site, http://java.sun.com.
The application is in a single JAR file and can be downloaded from here. After the installation of JAVA 1.4 and downloading the application open a shell (command prompt) window in the directory where you downloaded the "mubi.jar" file. At the command-line type the following command:
java -jar mubi.jar
On some operating systems (Windows, MacOS X) the "mubi.jar" can be started as a regular application in the file manager windows. This is due to the fact, that the Java 1.4 installation registers the ".jar" file type as an executable.
Current version of MUBI, v2.0, includes many score normalization and fusion algorithms. You may prefer to download a compressed repository (save as .ren file) that includes appropriate help files.
MUBI is an application for analyzing biometric systems. Only one biometric system can be analyzed at a time. Each system contains a number of biometric devices. The inputs for each device are the genuine and the imposter scores. When we want to add a device to the project (biometric system), we need to provide also two text files that contain the genuine and the imposter scores.
After we add devices to the project, the project with all the information about the devices can be saved into a single binary file.
For each device we can plot the estimated density curves for it's genuine and imposter scores.
The main window of the application allows only one biometric system (project) to be loaded. In this window we can add and remove devices, change their properties and save the system in a single file. Also, density plots and ROC curve plots can be started from this window.
The look of the entire application can be changed in the "Look and Feel" menu. The default look is the current system look.
Initially when the application is started it contains an empty project. But we can always start a new project by selecting "New Project" from the "File" menu.
Figure 1. Empty project
Devices can be added by selecting "Add New Device..." from the "Device" menu. After this, a dialog shows up where we can type the name of the device, set whether higher scores are better or lower scores are better, and select the genuine and the imposter score files with a click on the appropriate "Browse" button. The genuine and the imposter score files must be text files that contain only real numbers.
Figure 2. Adding a new device
After adding three devices the main window looks like this:
Figure 3. Three devices added to the project
Since the project now contains some number of devices, we may decide to modify some of its default properties. The first column tells the index of the devices in the project. Cells from this column are not editable. If we click on any of this cells, the entire row for that cell will be selected. The other cells in the rest of the columns are editable:
- Device name: By double-clicking any cell in this column we'll be able to edit the name of the device.
- Higher score - better: This column is used to tell us, whether higher scores for each device are better, or the lower ones. If we want to change this property, we can just click on the check box inside of the cell.
- Color: This column represents the color that will be used to plot device's ROC curve when plotting "k out of n" ROC curves. When clicking on this cell, a color dialog will show up to select a color for the device.
- Number of thresholds: Tells how many thresholds will be picked up between the min and max value for the thresholds, when plotting "k out of n" system ROC curve. The cell can be edited by double-clicking it.
- Min value for thresholds: By default, this value is the score with minimal value found from the genuine and imposter scores. Double-clicking any cell of this column will enable us to edit the value. If we want to check the ranges for the genuine and imposter scores on the currently selected device, we can go to the "Device" menu and select "
...". View Score Ranges
- Max value for thresholds: By default, this value is the score with maximum value found from the genuine and imposter scores. Double-clicking any cell of this column will enable us to edit the value. If we want to check the ranges for the genuine and imposter scores on the currently selected device, we can go to the "Device" menu and select "
...". View Score Ranges
Additional property that is not listed in the columns, is the line-style property for each device. This property enables us to select a different line-style when the device's ROC curve is plotted on the "k out of n" plot. After selecting the desired device, from the "Device" menu we select "Change Line Style" and a dialog will show up to select a line style:
Figure 4. Changing line style for a device
If we want remove some devices from the system, we can select the desired devices and then from the "Device" menu, select "Remove".
After editing the properties, the system may look like the following:
Figure 5. After all changes.
We can save the project using the "File" menu and selecting "Save Project" or "Save Project As".
Scores (genuine or imposter) can be exported to a text files. In order to see the scores for a currently selected device, from the "Device" menu we can select "Export Genuine Scores to a Text File" or "Export Imposter Scores to a Text File". The scores will be saved to the selected output filename.
This action can be performed after we select the devices for which we want to see the density estimations. We can select more than one device, and then from the "Device" menu select "Plot". For each selected device a new window will appear. More information in section 5.
At the bottom of the main window, there is an option to analyze a "k out of n" system. The number k can be entered, and then by pressing the "Plot System" button, a new window will appear (see section 6 for more details). While there are such windows opened, the data for the system in the main window cannot be modified, since it is required for drawing the "k out of n" windows. If we want to edit some data for the system (change device thresholds, add new device, remove devices, ...) we must close all "k out of n" windows.
Warning: This is a time consuming process and big numbers for the "Number of thresholds" of each device may result in more waiting. For example, for 3 devices and for each device the number of threshold set to 100, the process takes around 5 seconds on Intel Celeron 1.5GHz.
The following figure shows how a density estimation window looks like:
Figure 6. Density estimation for the device "Face"
For the imposter and genuine curves, we can change their color or line style. We can also choose whether any of the curves will be shown or not, by clicking on the appropriate "Show" check boxes.
For each curve, a bin width can be selected. Larger bin widths result in smoother curves. The boundaries of the possible bin widths depend on the size of the window. If the window is resized, then the boundaries of the bin widths will change because they depend on the available screen pixels to plot.
The density estimation plot, can be exported to an image. For the exported image, dimensions can be selected and the image type. The available image types may depend on the operating system that the application is running. Usually the types are: jpg, jpeg and png. Finally, the "Export to an image" button can be pressed to select the output file. On the right side of the output image, additional 200 pixels will be added to the width of the image for displaying the legend.
The following figure shows the "k out of n" plot window:
Figure 7. "k out of n" system
This window plots the ROC curve of the "k out of n" system and the ROC curves of each device. For the system's ROC curve and for its upper bound we can change the color and decide which of them will be visible. The devices are listed and for each of them, besides the name, their line-style and color are shown. By clicking on the check boxes, next to the device name, we can decide to show or not to show the device's ROC curve.
For each axis, the type of the axis can be changed to be linear or logarithmic.
When the mouse is moved over the graph, the FAR and GAR values are shown. If the current point is a point in the system's ROC curve, a label with show up saying "Click to select point" (see Figure 7, at the bottom of "Last selected point" area, below the table). After clicking the point in the table "Last selected point", the values of FAR, GAR will be shown including the threshold values of each device that formed this point.
The plot can be exported to an image in the same way like in the density estimation window.