Create Label(s)

Features:

• Enables user to label a single class with multiple labels. 

• Enables user to label multiple classes with a single label for each class.

• The labels can be saved to use for machine and deep learning applications implemented in IDCube.

Steps:

1.      Select Create Labels from Machine Learning Tab.

The image from the IMAGE DISPLAY panel (from the Main Interface) will automatically be shown.

2.      The Label maker toolbox offers two options:

a.      Label a single class (in addition, a background class is generated automatically, thus total two classes are generated)

b.      Label multiple classes (in addition, a background class is generated automatically, thus N+1 classes are generated).

Label a single class

1.      Click Label a Single Class pushbutton. The choice will be confirmed by the dropdown menu from the Selection Type. A Number of Labels window will be activated.

2.      Enter a number of labels you plan to have and click Apply. The background label will also be generated automatically.

3.      A pop-up will ask you to select the drawing tool. Click OK and select a tool (i.e., rectangle).

4.     Draw the number of areas that belong to the same class. The labels will be numbered such as shown below and automatically shown in the LABELS panel after the number of labels reaches the specified number (i.e., 3).

5.     Click Save Label(s) and select the directory. The default directory is the location of the original dataset. Notice: two files will be generated a png file and a m (or mat) file. Both files are necessary for machine and deep learning applications.

Label Multiple Classes

The steps are similar. In addition, you can use the Push Multiclass Label(s) function to visualize the selected labels on the left panel any time as long as the number of drawn labels is less than specified in the Number of Labels field.

Click Save Label(s) and select the directory. The default directory is the location of the original dataset. Notice: two files will be generated a png file and a m (or mat) file. Both files are necessary for machine and deep learning applications.