When you use images with large X/Y dimensions, e.g. large tile images, the segmentation will be only performed on a subset of the whole image in order to avoid long waiting periods. The current image subset maximum size in X/Y is 5000 pixels and is centered on the current view port. Nevertheless, all labels inside the complete image will be used for training, but the segmentation preview (pseudo-segmentation) will only be applied to that subset.

1 2 3
User Interface for Training

1

Image Labeling and Training

Here you can load the desired images for training, set up the required classes and select the options for segmentation.
You can also change the label opacity and the segmentation opacity there by adjusting the corresponding sliders. Opacity determines to what degree it obscures or reveals labels or segmentation. Opacity of 1% appears nearly transparent, whereas 100% opacity appears completely opaque. Additionally, you can hide all segmented pixels where the confidence value is below a certain threshold set by the Min. Confidence (%) slider.
With the different parameters of the Segmentation options and the Postprocessing options it is possible to further improve the results of the training and the (pseudo-)segmentation. Train & Segment starts the automatic training algorithm and then performs the pseudo-segmentation of the defined classes in the image.

2

Image Area
In the image area, the image selected in the left panel is displayed. You can label parts of the image as belonging to the class selected and highlighted in the Classes area on the left. When you are inside the image, the actual brush size for labeling is represented by a square. If the brush size is very small, the square is changed into a dotted circle with a small point inside.

3

Labeling Options
Here you can adjust certain parameters like Labeling Mode (Draw or Delete) or Brush Size. To switch the Labeling Mode, you can also use the shortcut Ctrl + D.