UNITED STATES: Meta, formerly known as Facebook, has introduced a cutting-edge artificial intelligence model that can identify specific objects within an image. The company has also unveiled an extensive dataset of image annotations, which it claims is the largest of its kind to date.
In a recent blog post, Meta’s research section said that it had created a powerful object identification model known as the Segment Anything Model. (SAM). The SAM model is designed to identify things in photos and videos, even if it hasn’t seen them before. Users can choose items by clicking on them or by typing text commands like “cat.” SAM effectively drew boxes around many cats in a picture during a demonstration in response to the instruction.
Meta has been using similar technology to SAM to do things like tag photos, moderate content that shouldn’t be there, and suggest posts to people on Facebook and Instagram. The release of SAM will make this advanced technology available beyond their internal operations.
The accompanying prototype and the SAM model are freely downloadable under a non-commercial license, but anyone who uploads their photographs there must agree to use the tool only for research.
Meta imagines SAM being utilised in a variety of domains where it is necessary to locate and segment every object in any image. For instance, it could be incorporated into larger AI systems for a more general multimodal understanding of the world, such as comprehending both the visual and text content of a webpage. In the AR/VR realm, SAM could enable selecting an object based on a user’s gaze and “lifting” it into 3D.
The tech giant says that SAM could help content creators because it would let them pull out parts of images to use in collages or video editing. The approach could also be helpful in a scientific study by enabling researchers to find and follow particular animals or things in video footage of natural occurrences on Earth or in space.
Also Read: Meta Announces Job Cuts: 10,000 Employees to be Sacked in the Year of Efficiency