Skip to content

GUI Trainer Application

The GUI trainer application allows you to train images and evaluate trackability scores via an easy-to-use GUI application.

Loading images

  1. Start the Track Target Trainer application, then click the Image Target button on the left.

    SelectImageTarget

    Tip

    Display Scaling

    You can change the display scaling of the Trainer application under Settings | Enable UI scaling in the top right corner of the start screen.

  2. Click the Load button on the top-left side, and select images to train.

  3. Once the images are loaded, set the size and name of the targets individually.

    Tip

    Batch Renaming You can also set target names in batch by clicking Batch Rename button on the top-left side.
    Using this function enforces the target names to be numbered consecutively, for example TARGET_001, TARGET_002 and so on.

    Info

    Any portrait images might be rotated by 90 degrees automatically because of the efficiency of the internal process.

Once the names and sizes are set, click the Import button on the top-right side.

Train your targets

  1. Select the image to be trained from the Image List.
    You might change name and size in the Property section on the right side. Click the Change button to confirm any changes.
    To select all targets at once, click Select All from the Edit toolbar on the left-top side.

  2. Click Train button on the top-right side to generate trackable data for all selected images.

  3. Once the training is complete, click the Done button.

Check the trackability score and delete targets

Once the training is complete, you can evaluate whether your image can be tracked robustly from the Trackability Score and the Visualized Heatmap.

  • Trackability Score
    The score can be found on the right side - from 0 (bad) to 100 (good). An ideal image also should have a high Trackability Score. A simple guideline for this is to have a score of more than 50. If your image has a low Trackability Score, consider changing to a different image.

  • Heatmap
    The Heatmap indicates which parts of the image contain good features. An ideal image for tracking will have a lot of violet or blue regions, and very few red regions. If your image does not contain many violet or blue regions, consider changing to a different image. To ensure robust tracking the feature-rich areas should be distributed evenly over the whole image.

Save trackable data

Once you have completed training, click the Save button on the top-right corner to save the generated Trackable Data.

Back to top