SAM is the image segmentation model with which Meta intends to democratize this technology

Madrid, April 5th. (Portaltic/EP) –

Goal Share your progress on image segmentation modellaying the foundations for a tool that helps recognize and distinguish objects, with applications in areas such as video editing or livestock.

The Meta project seeks to create an accurate segmentation model for specific tasks, and for this it has developed the Segment Anything Model (SAM) and its associated dataset (SA-1B), the largest segmentation dataset.

“SAM has learned a general idea of ​​what things are and It can generate masks for any object in any photo or videoeven for objects and image types for which it has not been trained, the Meta researchers explain.

You can also do interactive hash and auto hash, which makes it Flexible modelaccording to the company His blog on artificial intelligence.

The technology company aims for SAM to help democratize access to this kind of technology and that It can be applied to a variety of fields. For example, it can help in creative applications in video editing, but also in locating animals or tracking natural events on Earth.

This model, along with SA-1B, is available to researchers under an open source license (Apache 2.0).