Managing Ontologies for Class Tags

Nexus allows the adding of unique attributes of three types, Categorical, Numerical, and Text, to class tags. To create these attributes, one can select Manage Tags -> Advanced Tag Options, create a new class tag and click on it. To modify attributes of an existing class, simply click on the existing class tag. You will be brought to the menu below. On the left, you can change the tag properties such as default color of the tag, the type of annotation that is being made with the tag, the unique tag ID, and the tag description. On the right side, you can create new attributes with options to set an attribute name, the attribute type, select whether it is required or not, and include a default value if it is. Created attributes will be listed below, and you can edit them by clicking on the three dots -> Edit Attribute.

Attribute Types

Name	Description
Categorical	Choose a single value out of a list of options, e.g. car brands, dog breeds, etc.
Numerical	Supports both integer and floating point inputs.
Text	Supports freeform text input.

Setting Attribute Values

Once the attributes for each class tag have been saved, you can begin creating annotations. To set attributes for each annotation, select the created annotation and the Attributes tab on the right will appear.

If the values of any required attributes have not been set, the annotation will appear in red to indicate this. The asset cannot be submitted to the next stage of the annotation workflow until all annotations have fulfilled this criteria.

Once the values are selected, you will have to select Save button at the bottom to save the attributes to the annotation.

Ontology Applications

The following examples indicate the flexibility and practicality of Ontologies for various datasets.

Optical Character Recognition Dataset

With the example images shown above, we can make an OCR dataset by using the Rectangle tool to label lines of text and caption them with a string attribute.

Generative AI Dataset

For Generative AI datasets, it is important to know a few various aspects, such as the setup used to create such images, as this enables more pattern recognition to understand what is working best. This can be used to label ground truth images with their corresponding captions and also generated images with what was used to generate them.

Below, we have three attributes: Prompt, Negative Prompt, and Artistic Style categories used to augment the overall style of the generated image.

Here, we fill out the attributes and indicate the prompts, negative prompts, and artistic style used to create the image.

Common Questions

Can you create attributes with the same name across multiple classes?

Yes, as each attribute has its own attribute ID, and attributes are unique up to class tags.