Image Annotation

Oriented bounding boxes can be manually changed or added in the central view (6 in Fig. 13):

*Adding new boxes: *There are three steps to adding new boxes:

  1. First, double-click with the left mouse button on the position where you want the bottom edge of the bounding box to be. This border will always be green (Fig. 14).

  2. Without holding down the mouse button, click with the left mouse button on the upper edge of the object to set the height of the bounding box.

  3. To determine the width of the bounding box, click on any side point of the object with the left mouse button, again without holding down the mouse button.

The box label corresponds to the active class in the list (5 auf Fig. 13).

Selecting a bounding box: *Existing boxes can be selected with a simple left-click. A selected box is highlighted by a slightly different color and four yellow, circular corner handles (Fig. 14). *

Modification of existing boxes: As long as a bounding box is selected, you can change its label by selecting another label from the label overview or by typing a new label.

Correcting the bounding box: A selected box can be manipulated in several ways:

  • Drag the inside of an active box to move it.

  • Dragging the edges of an active box changes its size in that direction.

  • Drag the corners of an active box to rotate it around its center.

Deleting the bounding box: Right-click an active box to remove it.

Keyboard shortcut Operation

Ctrl/Shift + i/j/k/l

Move current box one/ten pixels in the corresponding direction

Ctrl/Shift + u/o

Rotate current box one/ten degrees counter/clockwise.

Ctrl/Shift + v/b

Increase/decrease current box height by one/ten pixels

Ctrl/Shift + n/m

Increase/decrease current box width by one/ten pixels

After each saved annotation, the LumiScan Annotation Framework also writes the following information to a JSON file:

  • User name

  • Unix timestamp

  • Image name (PNG)

  • Save file name (CSV)

The JSON file is saved in the raw directory under the name of the saved image. Sequential comments are appended so you can track who made a change and when.