Bounding Box
  • 15 May 2023
  • Dark
    Light
  • PDF

Bounding Box

  • Dark
    Light
  • PDF

Article Summary

Bounding Box

* Available on image and video items

Bounding boxes are generally used in object detection and localization tasks. They are relatively easy and quick to use.

Create a Bounding Box Annotation

recipe

  1. Go into the annotation studio.

  2. Select the Box tool.

  3. Draw the box on the canvas.
        A. 1-Click method – click and hold the mouse button on the point where you would like the top-left corner of the bounding box to be; drag the mouse to the bottom-right corner of the box; release the mouse button.
        B. 2-Click method – click once on the point where you would like the top-left corner of the bounding box to be; then click again on the bottom-right corner of the box.

  4. Drawing a box inside a box - While toggled ON, the bounding-box tool allows to either drag & drop existing boxes when clicking on them, or draw new box annotation when clicking anywhere else. Drawing a box inside an existing box annotation is available when holding down the SHIFT key and using the mouse to draw the new box.

  5. Use Pixel Measurement to create accurate bounding boxes measured by pixel size.
    This feature can be enabled or disabled by toggling ON/OFF the Pixel Measurement toggle in the Annotation Settings.

Editing Bounding Boxes

  1. The Box tool must be selected in order to edit bounding boxes.
  2. Click on the box you would like to edit to have it selected.
  3. Click and hold on the box and drag the cursor to move the box on the canvas.
  4. Click and hold any of the 8 points on the box's border and drag to resize the box.
  5. Hold SHIFT+Arrow-key to make fine position adjustments to the box location.

Box to Polygon & Box to Semantic Tools

Automatically convert a bounding box into a tight contour polygon around an object or a semantic layer.
Select a bounding box and activate box-to-polygon or box-to-sematic from the 3-dot action button above the annotations list.

OCR

OCR (Optical Character Recognition) mode requires the annotator to enter the text seen in the image (and bounded by the box/polygon).

OCR Mode must be enabled in the recipe instructions. Click here for more information.

  1. Draw a bounding box: a dialogue box will show.
  2. Enter the text recognized in the image and bounded by the box annotation and press ENTER.

Box Rotation

Box rotation mode allows the annotator to rotate the box in any direction. When enabling this option, the box annotation information includes the rotation angle.

  • Box-rotation mode must first be enabled from recipe settings. Check the option "Allow box rotation" in "Annotation Controls'' section.
  • Box rotation is applicable to image only, and not supported for video annotation.

What's Next