Step 2 - Label specs

The second step to creating a project is building a list of image categories, object classes and object properties (a.k.a. attributes).

Image Category

Step 1 - If you select Image Category as one of the annotation types, you will be asked to configure the project separate from object detection annotations.

1. Create a category (e.g. Time)

2. Create category properties (e.g. Afternoon, Evening)

3. Create a category group

๐Ÿ“ฝ๏ธ You can select whether to set Video Categories per frame. If you toggle Per Frame, you can label objects in different properties by each frame.

Object Detection

Step 1 - This is the page you will see if you selected Box, Polyline, Polygon Segmentation and/or Keypoint as the annotation type.

โ€‹

Object Class

1. Create the list of object classes by clicking + on the top of the Object Class card. You can start to designate all the relevant object classes for your current project.

2. Select the annotation type for each object class in the drop-down list.

Bounding Box
Polyline
Polygon
Keypoint
Category
Bounding Box
Polyline
Polygon
Keypoint
Category

You can create a Custom Keypoint Template as you like.

  • You can copy and paste the JSON file that contains the location, name and color information of the keypoints. Or, you can edit the JSON template directly on the web browser.

  • You can create multiple custom keypoint templates in a single project.

โ€‹

Class Property

You can set class properties to further define the object's characteristics.

1. Select the Object Class that you want to add a property then click + on the Class Property box

2. You can set the property types as Multiple Choice, Multiple Selection, and Free Response.

3. For Multiple Choice and Multiple Select types, you can set the property names and also set as a required property by clicking the Required toggle.

๐Ÿ“ฝ๏ธ On a video project, you can set the object properties per frame. If you toggle Per Frame, you can label different object property values per frame.

โ€‹

Configure Auto-Labeling

You can map the object classes with the Auto-Label object classes as following.

1. Open the Configure Auto-Label window by clicking the blue star located directly to the right of each object class.

2. Select the object class that you want to set Auto-Label.

3. Select all the classes on the check box that you want to label using Auto-Label AI.

4. Auto-Label configuration is done once you click Save on the top right corner.

โ— Availability of Auto-Label AI depends on the annotation type. You can find more details hereโ€‹

โ€‹

Setting Object Class Color

You can set the annotation color for each object class. Setting each class with different colors to distinguish each other can help improve labeling and validation efficiency, especially for complex projects with a lot of object classes.

1. Click the colored circle located on the left of the Object Class.

2. Select a color.

โ— You cannot change the colors of objects that are already labeled

โ€‹

Number of annotations per image

You can set the minimum/maximum number of how many annotations of each object class can exist in a label. This option is useful to avoid human errors when you know how many annotations should exist in a label. Assignees will not be able to submit a label if this requirement is not met.

1. Click the object class you want to set the min/max annotations per image

2. Enter the number on the field located on the bottom right corner of the page

  • Minimum : The default value is None. The value cannot be negative or higher than the maximum.

  • Maximum : The default value is None. The value cannot be negative or lower than the minimum.

โ€‹

Click Next on the left bottom if you are done.

โ€‹

Any other questions? E-mail us at [email protected].