Projects: Where Data Meets Models
A project in Datature Nexus is the container that holds your dataset, annotations, training runs, and deployed models. Each project maps to a single computer vision task. Keeping projects well-organized from the start makes it easier to track experiments, compare model versions, and share results with your team.
What This Tutorial Covers
Datature walks through project creation and configuration on Nexus:
- Creating a new project and naming it with a clear task description
- Selecting the annotation type (bounding box, polygon, keypoint, classification)
- Uploading or connecting assets from local files or cloud storage
- Understanding the project dashboard: assets, annotations, workflows, and artifacts
The full walkthrough runs about four minutes and covers the decisions you make at project creation time.
Structuring Projects for Long-Term Use
Teams that train multiple model versions or test different architectures benefit from a consistent project naming convention. One approach: include the use case, dataset version, and model type in the name. Datature supports running multiple training workflows inside a single project, so you can compare YOLOv8 against D-FINE without duplicating your dataset.
Once your project has data, the next step is annotation. See Introducing Advanced Search for Exploring and Managing Data for tips on navigating large datasets inside Nexus.
