The Segment Anything Model (SAM) is now available in image, tiled-image and video annotation tools. This integration helps project managers and contributors speed up the annotation process by automatically generating segmentation, bounding boxes and polygons.
With SAM you can significantly reduce manual effort and improve annotation consistency, particularly in large or complex visual datasets.
SAM can be used in any job using cml:shapes, cml:tiled_image_annotation and cml:video_shapes, talk to your Appen contact about how to enabling SAM for your jobs.
How it Works
When SAM is enabled a magic wand icon will appear in the job interface. Annotators can decide to use it or not at the unit level, by clicking or unclicking the icon.
- Contributors click on the icon and then an object in the job and SAM will automatically generate a precise outline around the object. If the job has an ontology, contributors will still need to assign the appropriate class to the shape.
- Contributors click on the icon again to de-activate the model, when the icon is greyed-out, contributors will manually create shapes.
