Data labelling using Graph UX

Data labelling is an expensive, time-consuming, and tedious process when building a machine learning model. To overcome this challenge in the machine learning world, Graph UX provides the functionality to collect and label the real-time data simultaneously, thus, saving time and efforts involved in labelling.

After you connect and set up the local microphone or serial capture to collect the data, you can utilize the same graph to label the real-time data, by adding a Pre-defined Labels node onto the canvas. Define the classes in the Predefined Labels node for one time and utilize the defined classes to label the data on a click of button.

Defining classes in the Predefined Labels node

To define classes in the Predefined Labels node, follow the steps:

  1. Double-click the Main.imunit where you created the graph for collecting data.
  2. Expand Projects and drag and drop the Predefined Labels unit onto the canvas.The Predefined Labels node displays the default classes.
  3. Define the classes in Predefined Labels node by entering every class in a new line.
  4. Navigate to the toolbar and click the Start button to open the session file (.imsession). An empty session file opens displaying the pre-defined classes in the Labels bar.

Real-time data labelling

To label the real-time data, follow the steps:

  1. Click the Record button to start the recording.
  2. Select a class in the Labels bar to enable that particular class. As you enable the class, the label with that class is added to the real-time data.
  3. Select the class again to disable that particular class. Similarly, you can enable or disable the other classes and label the data.

You can also enable or disable a class using the keyboard shortcuts displayed when hovering over the classes in the Labels bar.

  1. Click the Record button to stop the recording after you have labeled the data.
  2. Select File>Save Main.session As, if you only want to the save the session file. However, select File>Save to save the session file as well the data and the label track. The Save New Session window appears.
  3. In location, click the three dot and select the desired location to save the file.
  4. In Session Name, enter the name of the session file.
  5. In Session Name, enter the name of the session file.
  6. In Track options, select the save checkbox corresponding to the tracks you want to save. You can also edit the track name and file name by double clicking the respective columns.
  7. Click OK to save the files.
  8. Repeat the instructions from step 1 to create multiple session files.