Auto Annotation
Canvas Agent
Use models to speed up data annotation on the canvas
Canvas Agent Overview
The Canvas Agent provides an interactive way to auto-label specific frames directly on the Infinite Canvas, augmenting your manual annotation efforts with AI assistance.
Introduction
While we can create fully auto-annotated jobs as shown on Job Creation, we can also choose to auto-label specific frames directly on the Infinite Canvas.
Features and Usage
We’ve added a new “Agent” tab to the left sidebar which provides an agent-like interface where you can:
- Add labels and descriptions
- Customize confidence scores
- Test on a single frame
- Apply configuration to multiple frames in the current job
While we are confident in the performance of our flows and incorporated models, it’s recommended to review auto-generated annotations, as predictions are only as good as the confidence score and descriptions provided to labels of interest.
Best Practices
For optimal results:
- Always have a human review the automatically generated annotations
- Adjust auto-generated annotations when necessary to ensure quality
- Use confidence scores as a guide for prediction reliability