Segment Anything Model
A visual foundation model introduced by Meta, focused on image segmentation tasks. It forms an important part of Meta's AI technology stack.
Core Facts
Timeline (last 90 days)
The Segment Anything Model (SAM) has significantly improved general segmentation capabilities, but for semi-transparent details such as hair and wedding veils, dedicated matting models (such as ViTMatte) are required to achieve commercially viable accuracy.
All Facts (2)
SAM (Segment Anything Model) is a model open-sourced by Meta that achieves zero-shot segmentation of arbitrary objects through large-scale pre-training on 11 million images and 1 billion masks.
65%UnverifiedThe Segment Anything Model (SAM) has significantly improved general segmentation capabilities, but for semi-transparent details such as hair and wedding veils, dedicated matting models (such as ViTMatte) are required to achieve commercially viable accuracy.
50%