An ontology is a structured framework that defines all classes, attributes, and relations used to describe and organize labels in a project.
In a multi-level labeling structure, each class can have attributes and defined relations to other classes, such as a “vehicle” class with type, color, occlusion level, and motion state.
A well-designed ontology helps annotators capture rich, consistent, and semantically clear information to support model training and downstream reasoning.
Ontologies are usually defined before labeling starts, defined in the annotation guidelines, and updated as requirements change or new cases appear.
Xtreme1 by BasicAI was the first open-source multimodal training data platform to incorporate an ontology system.
An Ontology example with class and attributes
