All labeling activities in Label Studio occur in the context of a project. After you create a project and import data, set up the labeling interface and labeling configuration for your project. This setup process is essential to your labeling project.
Configure the labels and task type for annotators using the templates included with Label Studio or by defining your own combination of tags to set up the labeling interface for your project.
- Select a template from the available templates or customize one.
- Label Studio automatically selects the field to label based on your data. If needed, modify the selected field.
- Add label names on new lines.
- (Optional) Choose new colors for the labels by clicking the label name and choosing a new color using the color selector.
- Configure additional settings relevant to the labeling interface functionality. For example, when labeling text you might have the option to Select text by words.
- Click Save.
You can make changes to the labeling interface and configuration in the project settings.
You cannot remove labels or change the type of labeling being performed unless you delete existing annotations using those labels.
- In Label Studio UI, open the project you want to modify.
- Click Settings.
- Click Labeling Interface.
- Browse templates, update the available labels, or use the
Codeoption to further customize the interface using tags.
If you make changes to the labeling configuration, any tabs that you might have created in the data manager are removed.
You can customize a labeling config template or use a custom configuration that you create from scratch using tags. If you create a custom configuration that might be useful to other Label Studio users, consider contributing the configuration as a template.
The labeling configuration for a project is an XML file that contains three types of tags specific to Label Studio.
|Tag type||When to use|
|Object||Specify the data type and input data sources from your dataset.|
|Control||Configure what type of annotation to perform and how the results of annotation appear.|
|Visual||Define how the user interface looks for labeling.|
You can combine these tags to create a custom label configuration for your dataset.
For example, to classify images that are referenced in your data as URLs (
$image_url) into one of two classes, Cat or Dog, use this example labeling configuration:
<View> <Image name="image_object" value="$image_url"/> <Choices name="image_classes" toName="image_object"> <Choice value="Cat"/> <Choice value="Dog"/> </Choices> </View>
If you want to customize this example, such as to allow labelers to select both Cat and Dog labels for a single image, modify the parameters used with the Choices control tag:
<View> <Image name="image_object" value="$image_url"/> <Choices name="image_classes" toName="image_object" choice="multiple"> <Choice value="Cat"/> <Choice value="Dog"/> </Choices> </View>
You can configure your labeling configuration with the server API. See the Backend API documentation for more details.