Annotate images for use in your computer vision projects.
Roboflow Annotate provides a fast, robust interface through which you can annotate images.
You can annotate images using bounding boxes and polygons.
You can annotate images by:
Drawing bounding boxes and polygons manually
Using Label Assist, a tool that uses model checkpoints (i.e. a previous version of your model) to recommend annotations
Using Smart Polygon, a feature that creates polygon annotations with a few clicks.
Using Label Assist with SAM, which uses the Segment Anything Model to create more precise polygon annotations with a few clicks.
Bounding Boxes vs. Polygons
With the option between drawing bounding boxes and polygons, you may wonder: what is the difference between these two annotation types?
Bounding boxes -- boxes drawn around an object of interest in an image -- are easier to draw than polygons, thus taking up less annotation time. Polygons, on the other hand, are more precise, and may lead to a slight increase in performance.
For segmentation tasks, you need to annotate images with polygons, as you are training your model to segment specific items from an image with precision.
This section of the Roboflow documentation shows how to annotate images using each of the above methods.
Roboflow is working on Autodistill, an open source project that can help automatically label images for your computer vision project. Check out the Autodistill documentation for more information on how to get started.