Segment Anything Model (SAM)

Manually selecting and masking objects in images is tedious work — and most tools fall apart when they encounter something unusual. The Segment Anything Model (SAM), developed by Meta AI, is built to handle virtually any segmentation task, even ones it's never seen before. Give it a point, a box, or any prompt, and it generates high-quality object masks instantly.

SAM can also identify and segment all objects in an image without any prompting at all. Its key innovation — a promptable segmentation system — allows it to generalize to new tasks with zero-shot capability, making it a powerful foundation model for developers, researchers, and anyone working with visual content at scale.