AI Labeling

Use AI tools to speed up your labeling process.

AI Labeling features use credits at the rates listed on our credits page.

Roboflow offers four AI-powered labeling features to help speed up your Workflow:

Label Assist: Use a foundation model, a model you have already trained, or a public model to label the image you have open.
Smart Select: Use the Segment Anything Model in your browser to annotate polygon masks with a single click.
Box Prompting: Draw examples of an object in an image, then let Box Prompting find all other instances of that object in the image.
Auto Label: Use a foundation model (GPT-6 Astra, SAM 3, Gemini, or Claude Opus 5.5), a trained model, or a saved Workflow to auto-label all images in your dataset.

We recommend each solution in the following cases:

Label Assist: Ideal for labeling one image at a time, with a trained model or a foundation model.
Smart Select: Ideal for labeling the first version of your dataset, adding new annotations that Label Assist missed, or adding annotations for new classes to a dataset.
Box Prompting: Ideal if you have lots of similar objects to label in an image (i.e. if you are labeling individual screws on a tray with dozens of the same screw).
Auto Label: Ideal if you need to label common objects (i.e. a vehicle) in bulk.

You can find out more about each option on their respective documentation pages.

Last updated 1 day ago

Was this helpful?