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 model you have already trained to auto-label images in your dataset.

• Smart Polygon: 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 like Grounding DINO to auto-label all images in your dataset with a single or multiple prompts.

We recommend each solution in the following cases:

• Label Assist: Ideal for labeling data once you already have a trained model.

• Enhanced Smart Polygon with SAM: 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.