Not every team has the bandwidth to label thousands of images in-house. Datature's External Annotation Service lets you outsource labeling to a managed workforce while keeping your data, annotation schema, and quality controls inside Nexus. This tutorial walks through the full process.
What This Tutorial Covers
- Requesting external labeling from within a Datature project
- Defining annotation guidelines and quality requirements
- Scoping the labeling task (which assets, which classes, what annotation type)
- Monitoring progress and reviewing completed labels
- Accepting or rejecting annotations before they enter your training pipeline
When to Use External Labeling
Internal teams hit capacity limits. A research lab with two ML engineers cannot label 100,000 images alongside model development. The external service handles the volume while your team focuses on model architecture, training configuration, and evaluation. Data stays on Datature's infrastructure throughout the process, so there is no export-to-third-party risk.
Quality Control Built In
The service is not a black box. You set the annotation guidelines, review a sample of completed labels before accepting the full batch, and reject anything that does not meet your standards. Labels are delivered inside your existing project, ready for training.
